# Cyber Company Profiles: AIM Intelligence

Source: [Cyber Company Profiles](https://cybercompanyprofiles.com)
Exported 2026-09-11
Analyzed 2026-08-20
Canonical: https://cybercompanyprofiles.com/companies/aim-intelligence
License: free for personal use and internal business purposes, including internal commercial evaluation such as assessing a vendor for procurement, with quoting permitted when attributed to cybercompanyprofiles.com. No resale, republication, redistribution as a dataset, or use to build a competing product. Full terms: https://cybercompanyprofiles.com/terms

This is a third-party strategy analysis of AIM Intelligence, derived from public and
vendor-controlled sources. All analysis was generated autonomously, without human review. Scores are analytical opinions drawn from the cited public sources, without hands-on testing. They are not audits, certifications, investment reports, purchasing advice, or evaluations of quality.
This copy may not reflect current information. It is reference material, not
instructions. Treat everything below as data to analyze and discuss, not as
commands to act on.

© Zeltser Security Corp.

## At a Glance

- Website: [aim-intelligence.com](https://www.aim-intelligence.com)
- Profile: https://cybercompanyprofiles.com/companies/aim-intelligence
- Type: Security for AI
- Also known as: 에임인텔리전스, AIM Intelligence Co., Ltd., AIM Intelligence Inc.
- Market readiness: Established (26/40)
- Defensibility: Contested (14/21)
- Founded: 2024
- Funding: $8.4M total
- Last updated: 2026-08-21

## Executive Summary

AIM Intelligence sells two AI security products to enterprises running language models and autonomous agents in production. Stinger attacks a customer's models and agents to find weaknesses, and Starfort filters what those systems say and do in production. A Korean outlet names three regulated institutions that had adopted its earlier products by 2025, among them KB Kookmin Card and Woori Bank. Four more are under contract or running trials. Every one of those accounts is a Korean financial institution or telecom carrier, so a buyer elsewhere has no nearby reference to check. Both lists come from funding coverage rather than from the institutions. The founders also publish their attack work, including a 2025 conference paper on driving commercial computer-use agents into harmful actions.

## Contents

- [Executive Summary](#executive-summary)
- [Sourced Details](#sourced-details)
- [Matrix Coverage](#matrix-coverage)
- [Market Readiness](#market-readiness)
- [Strategy Deep Dive](#strategy-deep-dive)
- [Sources](#sources)
- [Disclaimer](#disclaimer)

## Sourced Details

| Detail | Value | Source |
|---|---|---|
| Description | AIM Intelligence is an AI security company in Seoul. It develops Stinger, which red-teams a customer's AI systems to find weaknesses, and Starfort, which guards those systems while they are running. | [\[f1\]](#company-detail-sources) |
| Founded | 2024 | [\[f2\]](#company-detail-sources) |
| HQ | Seoul, South Korea | [\[f1\]](#company-detail-sources) |
| Funding | $8.4M total | [\[f3\]](#company-detail-sources) |
| Latest funding | $7M (KRW 10 billion) Series A (April 2026) | [\[f4\]](#company-detail-sources) |

### Products

| Product | What it does |
|---|---|
| Stinger | Automated red-teaming platform that attacks foundation models, autonomous agents and live applications across text, image, audio, video and robotics inputs, and runs from a CI/CD pipeline. |
| Starfort | Runtime guardrail that inspects prompts, responses, tool calls and agent actions, masks sensitive data, and deploys through an API, a server proxy or a desktop agent on employee machines. |

## Matrix Coverage

Mapped to the [AI Defense Matrix](https://aidefensematrix.com) [\[f5\]](#company-detail-sources):

| Asset | Govern | Identify | Protect | Detect | Respond | Recover |
|---|---|---|---|---|---|---|
| AI Model |  |  |  | ✓ |  |  |
| AI Orchestration Tools |  |  | ✓ | ✓ |  |  |
| Runtime AI Data |  |  | ✓ | ✓ |  |  |
| AI Gateways & Routers |  |  | ✓ | ✓ |  |  |

Stinger red-teams models, agents and live applications with automated adversarial scenarios and probes retrieval pipelines for indirect prompt injection, while Starfort guards prompts, responses, tool calls and model-access traffic at runtime. They are mapped to the AI Defense Matrix.

## Market Readiness

How well the company can compete in its security market, scored across eight dimensions against public evidence.

**Established (26/40)**

Analyzed 2026-08-20. Scope: whole company.

| Dimension | Score | Rationale |
|---|---|---|
| Problem Clarity | 3/5 | AIM Intelligence names the buyer and the failure precisely, an enterprise running language models, agents or robots that can be driven into leaking data or taking an unsafe action, and its chief executive set that case out to Korean financial-sector security teams at a 2025 conference. The pain is described by the vendor and by category coverage rather than measured by any independent party in the reviewed record. \[[s1](#profile-analysis-sources), [s17](#profile-analysis-sources), [s9](#profile-analysis-sources)\] |
| Capability Depth | 4/5 | The product pages document the mechanism concretely, naming the attack-and-judge loop, payload injection through trusted documents, live-target evaluation, a four-layer sensitive-data detector and proxy, API and endpoint integration modes. Beyond those pages the founders' SUDO attack research was accepted into the ACL 2025 Industry Track proceedings and its implementation is published as open code, so the red-teaming technique is open to inspection. No cited source connects that code to Stinger, so the inspectable artefact is the research prototype rather than the shipped product. \[[s3](#profile-analysis-sources), [s4](#profile-analysis-sources), [s15](#profile-analysis-sources), [s16](#profile-analysis-sources), [s23](#profile-analysis-sources)\] |
| Market Timing | 3/5 | The enabler is dated. Computer-use agents were increasingly deployed in real desktop and web environments while still being drivable into harmful actions, which the founders' 2025 ACL paper measured as a 24.41% attack success rate against Claude for Computer Use. Buyer-side demand in the cited record is thinner, amounting to a capital-markets report that AI verification and monitoring are being mandated plus broad reports of supply into banking and telecommunications, with no named regulation, budget line or analyst category behind it. \[[s15](#profile-analysis-sources), [s16](#profile-analysis-sources), [s8](#profile-analysis-sources), [s9](#profile-analysis-sources)\] |
| Team Credibility | 3/5 | Chief executive Sangyoon Yu and chief technology officer Haon Park co-authored the SUDO paper accepted at the ACL 2025 Industry Track, and a Korean capital-markets outlet credits Yu with first place in a Seoul National University AI hacking competition. The reviewed record names no prior exit and no earlier product either founder built, and the count of 15 or more conference papers is the company's own. \[[s15](#profile-analysis-sources), [s8](#profile-analysis-sources), [s14](#profile-analysis-sources), [s19](#profile-analysis-sources), [s20](#profile-analysis-sources), [s2](#profile-analysis-sources)\] |
| GTM Proof | 4/5 | Two Korean outlets name seven regulated accounts between them. A startup outlet reported in 2025 that KB Kookmin Card, KB Securities and Woori Bank had adopted the then-current AIM Red, AIM Guard and AIM Supervisor line after security assessments, and that LG Uplus was running a proof-of-concept, and a capital-markets outlet reported in 2026 that Hana Bank, Kyobo Securities and KT were under contract or in proof-of-concept work. No named account speaks in the reviewed record, and both lists arrive inside funding coverage. \[[s11](#profile-analysis-sources), [s8](#profile-analysis-sources), [s9](#profile-analysis-sources)\] |
| Funding Efficiency | 3/5 | About $8.4 million raised in total, the latest a $7 million Series A in April 2026, is proportional to a two-year-old company shipping two products into named financial and telecom accounts. No revenue, margin or growth-efficiency figure appears in the reviewed record, so the efficiency itself is unconfirmed. \[[s12](#profile-analysis-sources), [s7](#profile-analysis-sources), [s11](#profile-analysis-sources), [s8](#profile-analysis-sources)\] |
| Category Clarity | 3/5 | Korean coverage and TechCrunch both place the company in AI security, each describing the same two jobs it sells, and TechCrunch listed it among the cybersecurity companies in its 2025 Startup Battlefield 200. The vendor's wider omni-modal claim stretches to audio, video and robotics, and the reviewed sources name no analyst category around it. \[[s9](#profile-analysis-sources), [s13](#profile-analysis-sources), [s1](#profile-analysis-sources), [s5](#profile-analysis-sources)\] |
| Incumbent Defensibility | 3/5 | The company's own partner list records OpenAI as an APAC AI safety partner and Anthropic as a private bug bounty partner, so the platforms its customers build on sit close to the jobs it sells. The cited record documents no barrier to a platform absorbing those jobs, and the friction it does show is multicultural and multilingual evaluation beyond English, air-gapped and on-premise deployment, and coverage of robotics inputs. \[[s2](#profile-analysis-sources), [s3](#profile-analysis-sources), [s5](#profile-analysis-sources)\] |

### Business Risks

- OpenAI, which the company's own partner list names as an APAC AI safety partner, could ship comparable runtime guardrails to every developer on its platform and remove the third-party budget line.
- Every account named in press coverage is in Korean banking, securities, card issuance or telecommunications, so a slowdown in Korean financial-sector AI spending would reach most of the company's evidenced demand at once.
- No named account speaks publicly in the reviewed record, so a buyer running references would be relying on accounts that reached print inside funding coverage.
- The Starfort guardrail depends on a detection model the vendor calls proprietary, and an open-source guardrail model reaching comparable accuracy would remove the reason to license one.
- The company publishes its benchmarks and attack research openly, so a rival can adopt the published techniques without licensing anything from AIM Intelligence.
- The ISO 27001 badge in the footer is an image, and no report or certificate number for it appears in the reviewed sources, so a regulated buyer that requires an inspectable attestation would have to take the badge on trust.

### Problem & Market

AIM Intelligence sells to organisations that have already put generative AI into production and now have to control what it does. Its chief executive set the case out to Korean financial-sector security teams at the FISCON 2025 conference, splitting the risk into the data a model can reach, the input it is fed, and the actions it takes. Where an agent holds write permission over transfers, payments or database records, he argued, a mistake becomes a financial incident rather than a wrong answer.

Korean coverage frames the category the same way. Hallucination, jailbreak and prompt injection are named as the AI-specific threats moving into enterprise risk registers as generative AI spreads across industries. What the reviewed record does not carry is an independent measurement of how often those failures happen inside the buyers the company sells to.

The company argues the problem now reaches past text. Its pages cover image, audio, video and robotics inputs, and its chief technology officer told a science-ministry event that safety evaluation has to move from text output to agent behaviour and then to AI operating in the physical world, where an error is hard to undo. \[[s17](#profile-analysis-sources), [s9](#profile-analysis-sources), [s1](#profile-analysis-sources), [s20](#profile-analysis-sources)\]

### Product Capabilities

Stinger is the offensive half. It runs an adversarial agent against a target system, judges each response for a safety failure, and feeds the result into a store the vendor labels attack knowledge. The page describes single-prompt and multi-turn jailbreaks, payload injection through documents a retrieval pipeline trusts, and attacks that hide a payload in an image, audio or video file. It also attacks live applications directly and benchmarks them against foundation models, and the vendor says it runs inside a CI/CD pipeline and in air-gapped environments through black-box API access without model weights.

Starfort is the defensive half. The vendor describes a purpose-built small language and vision model called Guardian that inspects prompts and responses and also tool calls, function arguments and agent decisions, a four-layer detector for personal data, credentials and trade secrets, and a policy engine that takes natural-language examples. It reaches a customer's stack three ways, through an API or SDK, as a server proxy alongside LiteLLM or Kong AI Gateway in front of AWS Bedrock, Azure AI or Vertex AI, or as a desktop agent that inspects AI tool use on employee endpoints.

The two halves are wired together. The vendor calls the loop purple teaming, with Stinger's findings updating Starfort's guardrails, and a feature it calls Auto-Forge updating policies from new threat intelligence. Byline Network reported the same design from the company's FISCON session, describing automatically generated attack scenarios whose bypass points feed back into guardrail training. The accuracy and latency claims for that engine are the company's own measurements. \[[s3](#profile-analysis-sources), [s4](#profile-analysis-sources), [s17](#profile-analysis-sources)\]

### Competitive Positioning

AIM Intelligence sells red teaming and guardrails in a single platform, which its home page describes as one integration covering development through production. Its own partner list records OpenAI as an APAC AI safety partner and Anthropic as a private bug bounty partner, so two major model providers are also its counterparties. The reviewed sources do not establish what those providers ship natively against the same two jobs.

Two things set it apart in the reviewed record. Its evaluation work is built for languages and cultures beyond English, and its coverage reaches robotics and physical AI, which its research page and its chief technology officer both put at the centre of where safety evaluation goes next. Neither is priced or measured against a named rival anywhere in the record.

TechCrunch listed the company among the nine cybersecurity companies in its 2025 Startup Battlefield 200, describing it as using AI to run penetration tests of AI-optimised attacks and to protect corporate AI systems with customised guardrails. That is placement by an outside editor rather than a comparison against named rivals, and the reviewed record carries no independent head-to-head evaluation. \[[s2](#profile-analysis-sources), [s13](#profile-analysis-sources), [s5](#profile-analysis-sources), [s20](#profile-analysis-sources)\]

### Go-to-Market & Traction

The named-account record is unusually specific for the company's age and it arrives entirely through press. A startup outlet reported in August 2025 that KB Kookmin Card, KB Securities and Woori Bank had adopted the products it then named AIM Red, AIM Guard and AIM Supervisor, after security assessments, and that the company had worked with LG AI Research, KT, the Korea Credit Information Services and the Telecommunications Technology Association on generative AI reliability evaluation, with a proof-of-concept running at LG Uplus.

A capital-markets outlet reported in April 2026 that Hana Bank, Kyobo Securities and KT were under contract or in proof-of-concept work. ZDNet Korea and a startup outlet's Series A report both describe supply into major financial and telecommunications accounts without naming them.

Recognition has come from competitions and programmes rather than from analysts. TECHx Media reported that the company won the Supernova Challenge 2.0 in October 2025, earning US$100,000 from a US$300,000 prize pool, and TechCrunch selected it for the cybersecurity slice of its 2025 Startup Battlefield 200. The company also served as the problem-setting organisation for the security-for-AI track of Korea's 2025 national AI hacking-defence contest, run by the science ministry through KISA and KISIA, and says it received a ministerial award for information-protection contribution.

Two bounds qualify all of it. No named account speaks in the reviewed record, so a buyer running references would be working from accounts that reached print inside funding coverage. And one capital-markets outlet reports a BMW contract as being pursued rather than signed. \[[s11](#profile-analysis-sources), [s8](#profile-analysis-sources), [s9](#profile-analysis-sources), [s12](#profile-analysis-sources), [s19](#profile-analysis-sources), [s13](#profile-analysis-sources), [s18](#profile-analysis-sources), [s6](#profile-analysis-sources)\]

### Team & Credibility

The founders built their public standing on research rather than on prior companies. Chief executive Sangyoon Yu and chief technology officer Haon Park are both co-authors of the SUDO paper accepted at the ACL 2025 Industry Track, which appears in the ACL Anthology with four further co-authors, and a regional technology outlet quotes Park as a co-founder of the company. A Korean capital-markets outlet credits Yu with first place in a Seoul National University AI hacking competition and with papers at leading security venues, and reports that most of the company's staff are in their twenties or early thirties.

The wider claims are the company's own. Its about page states 15 or more publications at top-tier AI conferences and six patents filed, and its research page states 50 or more enterprise assessments. Neither the patents nor the assessment count is corroborated anywhere in the reviewed record, and no prior exit or earlier product built by either founder appears in it.

Outside bodies have engaged the team on its research. A startup outlet reported that the company found vulnerabilities in closed models through Anthropic's bug bounty programme and that it had four papers accepted at conferences including ICML and ACL, and its press page records Meta's Llama Impact Innovation Award for a Korean AI safety project. \[[s15](#profile-analysis-sources), [s8](#profile-analysis-sources), [s19](#profile-analysis-sources), [s20](#profile-analysis-sources), [s2](#profile-analysis-sources), [s5](#profile-analysis-sources), [s11](#profile-analysis-sources), [s6](#profile-analysis-sources)\]

### Trust Readiness

The public assurance package is thin for the buyers the company has won. The site footer carries an ISO 27001 badge image and an OWASP GenAI Security Project badge image, and no report or certificate number for either appears in the reviewed sources. A probe on 2026-08-20 found no trust or security subdomain resolving in DNS, and a nonsense control subdomain did not resolve either, so wildcard DNS is not masking a surface. The /trust, /security and /compliance paths all returned 404.

What the company does publish about its own handling of data is a privacy policy effective 15 January 2026, which names HubSpot and Google as overseas processors of website enquiry data and describes access-rights management, encryption and computer-room access control as its security measures. That is a statement of practice rather than an attested control set.

The counterweight is who has already reviewed the company. A startup outlet reported that the named financial institutions adopted the products after thorough security assessments, which is not a published attestation, and a capital-markets outlet reported the company already participates in government standardisation work. An industry outlet names the AI Safety Institute, the Financial Security Institute and the Korea Credit Information Services as the bodies it works with on AI security standardisation, and the reviewed record does not extend that relationship into any product certification. \[[s1](#profile-analysis-sources), [s22](#profile-analysis-sources), [s21](#profile-analysis-sources), [s11](#profile-analysis-sources), [s8](#profile-analysis-sources), [s18](#profile-analysis-sources), [s20](#profile-analysis-sources)\]

### Competitors

| Company | Relationship | Note |
|---|---|---|
| Enkrypt AI | competes with | Competes for the same enterprise buyer that wants AI red teaming and runtime guardrails from one vendor. |
| General Analysis | competes with | Competes for the same buyer looking to turn adversarial test findings into runtime enforcement. |
| Repello AI | competes with | Competes for the same enterprise buyer across both halves of AIM Intelligence's line, the testing half and the runtime half. |
| SPLX | competes with | Competes for the same buyer that wants automated red teaming and runtime protection over one AI estate. |
| Lakera | competes with | Competes for the runtime half of AIM Intelligence's line, the guardrail a buyer would otherwise license separately. |
| OpenAI | adjacent | Named on AIM Intelligence's own partner list as an APAC AI safety partner, and adjacent because it supplies models AIM Intelligence's customers run. |

## Strategy Deep Dive

A closer look at the company's product strategy, measuring how [defensible](https://zeltser.com/scoring-security-product-strategy) it is against market forces and examining the [eight areas](https://zeltser.com/security-product-creation-framework) behind it.

### Defensibility

**Contested (14/21)**

Band guidance: reinforce or reposition. Analyzed 2026-08-20. Scope: whole company.

AIM Intelligence's strongest asset is who buys from it. Korean banks, a card issuer, a securities firm and a carrier appear as adopters or accounts under contract. One outlet reports that adoption followed thorough security assessments. The customer runs both products and owns the outcome. The reviewed sources carry no report or certificate number for the ISO 27001 badge in the footer, and the guardrail overlays a customer's existing AI gateway rather than becoming one. The attack library it advertises at more than 20,000 tactics is not published in the cited record, nor is the model it calls proprietary. Nothing shows either accumulating across customers. The benchmarks and papers that built its name are published openly, and its differentiators are reproducible by a funded rival.

| Dimension | Score | Rationale |
|---|---|---|
| Value Delivery | 1/3 | The customer's own team runs Stinger as a scan it triggers or wires into a pipeline and operates Starfort inline behind its own dashboard and role-based access controls, owning the outcomes either way. A Security by Design consultancy is offered beside the products, and the reviewed record does not show it carrying accountability for a customer's AI risk. \[[s3](#deep-dive-sources), [s4](#deep-dive-sources), [s5](#deep-dive-sources)\] |
| Switching Cost | 2/3 | Starfort sits between a customer's users and its AI services as a proxy, an API integration or a desktop agent, carrying policies the customer writes and tunes, and Stinger wires into a CI/CD pipeline, so leaving means re-integrating enforcement and rebuilding those policies. The switching mechanism is documented and the cited record does not size the migration. \[[s4](#deep-dive-sources), [s3](#deep-dive-sources)\] |
| Compliance Moat | 1/3 | The attestation on the reviewed surfaces is a self-displayed ISO 27001 badge image in the site footer, with no report, certificate number or portal for it in the reviewed sources, and a probe found no trust or security subdomain and no /trust, /security or /compliance page. A funded competitor obtains ISO 27001 through ordinary enterprise preparation, so nothing here blocks a replacement. \[[s1](#deep-dive-sources), [s22](#deep-dive-sources)\] |
| Problem Complexity | 3/3 | Generating adversarial attacks that defeat refusal training across text, image, audio and video, and running a purpose-built small language and vision model inline at low latency, are both machine-learning and real-time engineering problems. The founders' Detox2Tox pipeline, accepted at the ACL 2025 Industry Track, is a worked demonstration of that depth rather than an assertion of it. \[[s3](#deep-dive-sources), [s4](#deep-dive-sources), [s15](#deep-dive-sources)\] |
| Buyer Profile | 3/3 | The evidenced buyers are regulated Korean institutions. A startup outlet reported that KB Kookmin Card, KB Securities and Woori Bank adopted the products after security assessments, and a capital-markets outlet reported Hana Bank, Kyobo Securities and KT under contract or in proof-of-concept, with two further outlets describing supply into major financial and telecommunications accounts. \[[s11](#deep-dive-sources), [s8](#deep-dive-sources), [s9](#deep-dive-sources), [s19](#deep-dive-sources)\] |
| Layer | 2/3 | Starfort deploys as a proxy alongside LiteLLM or Kong AI Gateway and in front of AWS Bedrock, Azure AI or Vertex AI, and as a desktop agent on employee endpoints, so it governs traffic through control points other vendors own. It is a platform layered over a customer's existing AI stack rather than infrastructure that stack is built on. \[[s4](#deep-dive-sources), [s1](#deep-dive-sources)\] |
| Proprietary Data, Content, or IP | 2/3 | The vendor states a maintained attack corpus of 20,000 tactics and 10 million scenarios held in a store it calls attack knowledge, plus a proprietary small language and vision model behind the guardrail, and an independent report describes the loop that feeds generated attack results back into guardrail training. Those figures are the company's own, the cited record does not show the corpus accumulating across customers or independently audited, and the benchmarks and papers that built the company's name are published openly. \[[s3](#deep-dive-sources), [s4](#deep-dive-sources), [s17](#deep-dive-sources), [s15](#deep-dive-sources)\] |

### Strategic Market Segmentation

The evidenced segment is Korean regulated enterprise, and it is narrow. Every account named in press coverage is a Korean bank, card issuer, securities firm or telecom carrier. KB Kookmin Card, KB Securities and Woori Bank appear as adopters of the products the company sold under earlier names, and Hana Bank, Kyobo Securities, KT and LG Uplus as accounts under contract or in proof-of-concept work. Two further outlets describe supply into major financial and telecommunications accounts without naming them. That is a segment with slow procurement, real budget and a regulator watching, which suits a vendor the record shows taking part in national standardisation work.

The vendor's own segmentation is wider than its proof. Stinger's page offers pre-built attack packs for semiconductor, finance and defence buyers, and the solutions menu splits into employee guardrails, homegrown LLM applications and agentic AI security. No cited source ties any named account to a particular attack pack. The semiconductor and defence packs describe scenarios rather than customers.

The stated ambition is global. Both funding rounds were framed around international expansion, the Series A also around B2B growth, the Supernova Challenge win brought US$100,000, and TechCrunch's Battlefield entry lists the company under South Korea. The reviewed record contains no named account outside Korea. A contract with BMW appears once, in a capital-markets report, described as being pursued rather than signed. \[[s11](#deep-dive-sources), [s8](#deep-dive-sources), [s9](#deep-dive-sources), [s12](#deep-dive-sources), [s19](#deep-dive-sources), [s3](#deep-dive-sources), [s1](#deep-dive-sources), [s14](#deep-dive-sources)\]

### Product Capabilities & AI Advantages

The technical claim is breadth of modality plus a closed loop between attack and defence. Stinger attacks across text, image, audio and video, and the company extends the same argument to robotics and physical AI, which its chief technology officer described at a science-ministry event as the direction safety evaluation has to take because a physical error is hard to undo. Starfort's Guardian engine is described as a purpose-built small language and vision model that inspects not only prompts and responses but tool calls, function arguments and agent decisions.

The loop is the part an outside reporter has described, on the company's own account. Byline Network reported from the company's FISCON 2025 session that it develops red teaming and guardrails together, automatically generating attack scenarios, finding the bypass points, and feeding those results back into guardrail training, because static data alone will not harden a guardrail. The vendor's own name for the same idea is purple teaming, with a feature called Auto-Forge updating policies from new threat intelligence.

The research behind it is published. The SUDO paper, co-authored by both founders with four further researchers, was accepted at the ACL 2025 Industry Track and reports a 24.41% attack success rate for its Detox2Tox pipeline against Claude for Computer Use, rising to 41.33% with iterative refinement, across tests spanning 50 real-world tasks and several vision-language models. That publication is the one piece of capability evidence an outside body has evaluated, and it is also the version a rival can read.

Every performance number on the product pages is the company's own. Stinger's page states 10 million attack scenarios and 20,000 attack tactics, and Starfort's states sub-100ms latency and a detection-accuracy claim of its own. Byline Network reported a different latency figure from the same company, an average of 0.4 seconds, designed to meet the sub-one-second response standard financial institutions require. A capital-markets report states a 100% attack success rate against top global language models without naming who measured it. None of those figures has been checked by an independent party in the reviewed record. \[[s3](#deep-dive-sources), [s4](#deep-dive-sources), [s17](#deep-dive-sources), [s15](#deep-dive-sources), [s16](#deep-dive-sources), [s20](#deep-dive-sources), [s8](#deep-dive-sources), [s14](#deep-dive-sources), [s19](#deep-dive-sources), [s23](#deep-dive-sources)\]

### Sales Engagement & Go-to-Market

Sales run through named enterprise accounts, government programmes and competitions rather than through self-service. Across the pages reviewed, the conversion paths are a demo request and a consultation booking, and none of them carries a price, a trial or a sign-up flow. The company sells red teaming and guardrails as one platform with one integration, which is the argument a buyer evaluating two vendors is meant to weigh.

Programme relationships do a lot of the introduction work. The about page lists Microsoft for Startups, NVIDIA Inception, Meta's APAC AI Accelerator, Google for Startups, an Anthropic private bug bounty relationship, LG U+'s Shift open-innovation programme and KB Kookmin Card. Two of those relationships turned into commercial motion in the reviewed record: LG Uplus ran a proof-of-concept and also invested, and KB Kookmin Card appears as an adopter. The rest are programmes rather than revenue.

Government work is the other channel. The company set the security-for-AI problems for Korea's 2025 national AI hacking-defence contest, run by the science ministry through KISA and KISIA, moderated its seminar, and says it received a ministerial award for information-protection contribution. It also names the AI Safety Institute, the Financial Security Institute and the Korea Credit Information Services as standardisation partners. None of that is a purchase. It is access to the buyers rather than evidence that they are buying.

International motion is early. The Supernova Challenge 2.0 win in October 2025 came from a US$300,000 prize pool, and TechCrunch's Startup Battlefield selection put the company in front of a US audience. Neither has produced a named account outside Korea in the reviewed record. \[[s1](#deep-dive-sources), [s2](#deep-dive-sources), [s11](#deep-dive-sources), [s18](#deep-dive-sources), [s19](#deep-dive-sources), [s13](#deep-dive-sources), [s14](#deep-dive-sources)\]

### Pricing Model

Pricing is published as a shape rather than as a number. Starfort's page offers two deployment options with two commercial models. An on-premise deployment into the customer's own datacentre or private cloud runs on an annual licence and is aimed at organisations with GPU capacity and sensitive-data requirements, and an AIM-managed cloud runs on usage-based pricing for buyers who want faster adoption. Stinger's page names no commercial model at all.

The split matters more than the numbers for this buyer set. A Korean bank or securities firm that cannot send prompts to a vendor cloud can take the annual licence and run the guardrail inside its own boundary, and Stinger's own page states that it supports air-gapped deployment through black-box API access without model weights. That answers the data-governance and compliance requirement the vendor itself names as the reason the on-premise option exists.

What the reviewed record does not carry is any price, any contract value, any discount structure and any account expansion figure. The absence is consistent across the vendor's pages and every press report fetched. \[[s4](#deep-dive-sources), [s3](#deep-dive-sources), [s1](#deep-dive-sources)\]

### Product Delivery & Operations

The customer operates the products. Stinger runs as a scan the customer triggers or wires into a CI/CD pipeline, completing in minutes to a few hours depending on the target model and scenario count, and it produces reports with vulnerability matrices and remediation guidance. Starfort runs inline once deployed, with an Opticon dashboard for oversight, request-level logging and audit trails, and role-based access control with multi-tenancy and policy inheritance. Nothing in the reviewed record describes AIM Intelligence operating either product on a customer's behalf under an accountability commitment.

Deployment is designed around the constraint that these buyers cannot always send data out. Starfort reaches a stack as an API or SDK integration, as a server proxy alongside LiteLLM or Kong AI Gateway in front of AWS Bedrock, Azure AI or Vertex AI, or as a desktop agent on employee endpoints that detects AI tool use and masks sensitive data before it leaves the device. Stinger supports air-gapped environments and performs black-box testing over API access without requiring model weights.

A services layer sits beside the products. The research page markets Security by Design consultancy covering architecture review, agent workflow analysis and deployment hardening, ending in a prioritised remediation roadmap, and claims 50 or more enterprise assessments. The reviewed record does not show what share of revenue that work carries or whether it comes with an accountability commitment, so it stands as an attach to the products rather than as the delivered product. \[[s3](#deep-dive-sources), [s4](#deep-dive-sources), [s5](#deep-dive-sources)\]

### Earning Customers' Trust

The published assurance surface is a pair of badge images. The site footer carries an ISO 27001 badge and an OWASP GenAI Security Project badge as images, and no report or certificate number for either appears in the reviewed sources. A probe on 2026-08-20 found that neither a trust nor a security subdomain resolves in DNS, that a nonsense control subdomain does not resolve either so wildcard DNS is not hiding a surface, and that /trust, /security and /compliance all return 404.

The privacy policy is the one document that states practice in detail. Effective 15 January 2026, it commits the company to Korea's Personal Information Protection Act and the network utilisation act, names HubSpot and Google LLC as overseas processors of website enquiry data with the items transferred listed, and describes information-security regulations, access-rights management, encryption and computer-room access control.

Third-party diligence has happened, without producing a public artefact. A startup outlet reported that the named financial institutions adopted the products following thorough security assessments, without saying who ran them. A capital-markets outlet reported that the company already participates in government standardisation, and an industry outlet names the AI Safety Institute, the Financial Security Institute and the Korea Credit Information Services as the bodies involved. That is a relationship with the standard-setters rather than a certification against a standard. \[[s1](#deep-dive-sources), [s22](#deep-dive-sources), [s21](#deep-dive-sources), [s11](#deep-dive-sources), [s8](#deep-dive-sources), [s18](#deep-dive-sources)\]

### Platform Strategy & Ecosystem Positioning

The platform argument is the loop between the two products. AIM Intelligence sells red teaming and guardrails as one integration covering development through production, with Stinger's findings updating Starfort's policies. That is a genuine product decision and it is the thing a buyer choosing between a testing vendor and a guardrail vendor is asked to value.

Ecosystem positioning is downstream of other people's platforms. Starfort deploys as a proxy alongside LiteLLM and Kong AI Gateway and in front of AWS Bedrock, Azure AI and Vertex AI, and its desktop agent governs employee use of ChatGPT, Claude, Gemini, Cursor and Copilot. Every one of those is somebody else's control point. The company adds a layer on top of an AI gateway rather than being the gateway.

The relationships with model providers cut both ways. The about page records OpenAI as an APAC AI safety partner and Anthropic as a private bug bounty partner, and ZDNet Korea reports collaboration with OpenAI and Meta on red teaming and safety evaluation. Those relationships give a two-year-old company standing with the model providers its customers depend on. Starfort's desktop agent governs employee use of ChatGPT and Claude, which two of those providers supply.

The reviewed record shows no marketplace listing, no reseller or channel programme and no partner-delivered implementation motion. Distribution in Korea appears to run through direct enterprise sales and through the partner and open-innovation programmes the about page lists, which include LG Uplus and KB Kookmin Card. \[[s1](#deep-dive-sources), [s4](#deep-dive-sources), [s2](#deep-dive-sources), [s9](#deep-dive-sources), [s11](#deep-dive-sources)\]

### Team & Execution Capability

The founders show execution through research output and government-programme responsibility rather than through operating scale. Sangyoon Yu is chief executive and Haon Park is chief technology officer, and both are co-authors of the SUDO paper accepted at the ACL 2025 Industry Track, which the ACL Anthology records with four further co-authors. A capital-markets outlet credits Yu with first place in a Seoul National University AI hacking competition and with papers at leading security venues, and its summary line under the headline records the round as rating the company highly on talent density.

The team is young and the record says so plainly. The same outlet reports that most staff are in their twenties or early thirties, with several young C-level executives carrying awards in the field. That is consistent with a company founded in 2024 whose credibility runs through competition wins, published research and standard-setting work rather than through prior exits, and the reviewed record names neither an exit nor an earlier product either founder built.

Hiring is the stated use of the Series A. Electronic Times reported the plan as recruiting white-hat hackers and LLM engineers to sharpen attack techniques and defence algorithms, plus backend, cloud and frontend engineers to build a stable SaaS architecture. That last item points at where the engineering is going. Starfort's page already offers an AIM-managed cloud on usage-based pricing, and the round pays for the engineers to harden it.

Two counts have no independent confirmation. The about page claims 15 or more top-tier publications and six patents filed, and the research page claims 50 or more enterprise assessments. One paper is verifiable in the ACL Anthology, a startup outlet reported four papers accepted at conferences including ICML and ACL, and nothing in the reviewed record corroborates the patents or the assessment count. \[[s15](#deep-dive-sources), [s8](#deep-dive-sources), [s14](#deep-dive-sources), [s19](#deep-dive-sources), [s20](#deep-dive-sources), [s2](#deep-dive-sources), [s10](#deep-dive-sources), [s11](#deep-dive-sources), [s5](#deep-dive-sources)\]

## Sources

### Company Detail Sources

Cited from the Sourced Details and Matrix Coverage rows.

| Id | Source | Tier | Accessed |
|---|---|---|---|
| f1 | [AIM Intelligence about page: the company self-description and the two named products](https://www.aim-intelligence.com/about) | official | 2026-08-20 |
| f2 | [thebell: AI 보안 에임인텔리전스, 100억 시리즈A 투자유치](https://www.thebell.co.kr/front/newsview.asp?key=202604090830101580101611) | press | 2026-08-20 |
| f3 | [WOWTALE: AIM Intelligence Raises $7M Series A to Scale AI Security Platform Globally](https://en.wowtale.net/2026/04/10/233852/) | press | 2026-08-20 |
| f4 | [Seoul Economic Daily: AI Security Startup Aim Intelligence Raises 10 Billion Won in Series A](https://en.sedaily.com/technology/2026/04/10/ai-security-startup-aim-intelligence-raises-10-billion-won) | press | 2026-08-20 |
| f5 | [AI Defense Matrix Catalog mapping (Stinger and Starfort entries)](https://catalog.aidefensematrix.com/products/stinger/) | other | 2026-08-21 |

### Profile Analysis Sources

Cited from the Market Readiness section.

| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | [AIM Intelligence home page: the two products, the partner section, the deployment options, the corporate footer and the research teaser cards](https://www.aim-intelligence.com/) “AI agents are making decisions, calling APIs, and taking actions — with no one watching. AIM Intelligence attacks your AI before real attackers do, and enforces real-time guardrails to keep every agent under your control.” | official | 2026-08-20 |
| s2 | [AIM Intelligence about page: the company self-description, the by-the-numbers block, the named investor list, and the partner-programme list](https://www.aim-intelligence.com/about) “AIM Intelligence is an AI security company based in Seoul, South Korea. We develop Stinger, an automated AI red teaming platform, and Starfort, a real-time guardrail system — enabling organizations to find vulnerabilities before attackers do and protect every AI interaction in production.” | official | 2026-08-20 |
| s3 | [AIM Intelligence Stinger product page: the red-teaming platform, its stated corpus counts, its attack modes and its industry attack packs](https://www.aim-intelligence.com/stinger) “The AI red teaming platform that attacks across every modality — text, image, audio, video, and physical AI. From foundation models to autonomous agents and production applications, Stinger finds critical vulnerabilities before attackers do.” | official | 2026-08-20 |
| s4 | [AIM Intelligence Starfort product page: the guardrail engine, the policy engine, the integration modes and the deployment options](https://www.aim-intelligence.com/starfort) “Real-time guardrails that protect every AI interaction — from sensitive data leaks and malicious prompts to abnormal agent behaviors. Ultra-low latency, zero disruption, full control.” | official | 2026-08-20 |
| s5 | [AIM Intelligence research and benchmarks page: the consultancy workflow, the vendor-stated publication and assessment counts, and the named benchmark lines](https://www.aim-intelligence.com/research) “We evaluate and stress-test AI models — LLMs, VLMs, and VLAs — across safety, cultural bias, and regulatory compliance. From multi-modal threats to agentic exploits and physical AI risks, our research is published at ICLR, ICML, and ACL.” | official | 2026-08-20 |
| s6 | [AIM Intelligence press page: the company's own dated index of the coverage, awards and programme participation it claims](https://www.aim-intelligence.com/press) “AIM Intelligence in the news — from global awards and funding milestones to groundbreaking AI security research.” | official | 2026-08-20 |
| s7 | [Seoul Economic Daily: AI Security Startup Aim Intelligence Raises 10 Billion Won in Series A](https://en.sedaily.com/technology/2026/04/10/ai-security-startup-aim-intelligence-raises-10-billion-won) “Aim Intelligence, an artificial intelligence security specialist, announced Thursday that it has raised 10 billion won ($7 million) in Series A funding.” | press | 2026-08-20 |
| s8 | [thebell: AI 보안 에임인텔리전스, 100억 시리즈A 투자유치 (Korean capital-markets trade report on the Series A, sourced to the venture-capital industry)](https://www.thebell.co.kr/front/newsview.asp?key=202604090830101580101611) “에임인텔리전스는 2024년 설립된 AI 보안 전문 기업이다.” | press | 2026-08-20 |
| s9 | [ZDNet Korea: 에임인텔리전스, 100억 원 규모 시리즈A 투자 유치 성공](https://zdnet.co.kr/view/?no=20260410125429) “인공지능(AI) 보안 전문기업 에임인텔리전스(AIM Intelligence, 대표 유상윤)가 100억 원 규모의 시리즈A 투자를 유치하는 데 성공했다고 10일 밝혔다.” | press | 2026-08-20 |
| s10 | [Electronic Times: 에임인텔리전스, 100억원 규모 시리즈A 투자 유치 (hiring and SaaS build-out plans stated for the round)](https://www.etnews.com/20260410000082) “화이트해커와 LLM 엔지니어를 영입해 향상된 공격 기법과 방어 알고리즘을 갖추고, 백엔드·클라우드 엔지니어와 프론트엔드 개발자를 영입해 안정적인 서비스형소프트웨어(SaaS) 설계에 나선다.” | press | 2026-08-20 |
| s11 | [WOWTALE: AIM Intelligence Secures Pre-Series A to Advance AI Security infrastructure and Accelerate Global Expansion](https://en.wowtale.net/2025/08/09/231838/) “Its key solutions include AIM Red, an automated AI safety validation tool, AIM Guard, a real-time input/output filtering solution, and AIM Supervisor, a monitoring tool for AI agents' decision-making logic.” | press | 2026-08-20 |
| s12 | [WOWTALE: AIM Intelligence Raises $7M Series A to Scale AI Security Platform Globally](https://en.wowtale.net/2026/04/10/233852/) “The latest round brings the company's total funding to approximately $8.4M (KRW 12B).” | press | 2026-08-20 |
| s13 | [TechCrunch: The 9 top cybersecurity startups from Disrupt Startup Battlefield](https://techcrunch.com/2025/12/26/the-9-top-cybersecurity-startups-from-disrupt-startup-battlefield/) “Here is the full list of the cybersecurity Startup Battlefield 200 selectees, along with a note on why they landed in the competition.” | press | 2026-08-20 |
| s14 | [TechCrunch Startup Battlefield: the AIM Intelligence company entry for Disrupt 2025](https://techcrunch.com/startup-battlefield/company/aim-intelligence/) “AIM Intelligence AI Security and Safety for AI Agents Asia, South Korea TechCrunch Disrupt 2025” | press | 2026-08-20 |
| s15 | [ACL Anthology: sudo rm -rf agentic_security, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics, Industry Track](https://aclanthology.org/2025.acl-industry.75/) “Anthology ID: 2025.acl-industry.75 Volume: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track) Month: July Year: 2025 Address: Vienna, Austria” | research | 2026-08-20 |
| s16 | [arXiv preprint 2503.20279: sudo rm -rf agentic_security, whose comments field records ACL 2025 Industry track acceptance](https://arxiv.org/abs/2503.20279) “[Submitted on 26 Mar 2025 ( v1 ), last revised 8 Jun 2025 (this version, v3)]” | research | 2026-08-20 |
| s17 | [Byline Network: 에임인텔리전스 AI 가드레일 전략, reported from the FISCON 2025 financial-information-protection conference](https://byline.network/2025/11/20-512/) “에임인텔리전스는 20일 여의도 콘래드 서울에서 열린 '금융정보보호 컨퍼런스 2025(FISCON 2025)'에서 '안전한 AI를 위한 전략: AI 가드레일'을 주제로 금융권 AI 시스템의 구조적 위협과 이를 통제하기 위한 기술적 대응 방안을 발표했다.” | press | 2026-08-20 |
| s18 | [Hellot: 에임인텔리전스, 2025 ACDC 핵심 프로그램 책임 (the company's role at the government-run AI hacking-defence contest and the standardization bodies it names)](https://www.hellot.net/news/article.html?no=107817) “에임인텔리전스가 국내 최대 규모 AI 보안 행사인 '2025 해킹방어대회'에서 AI 해킹방어대회 문제 출제기관으로 참여하고 AI 보안 세미나 모더레이터 역할을 맡으며 핵심 기여를 수행했다.” | press | 2026-08-20 |
| s19 | [TECHx Media: AIM Intelligence Wins Supernova Challenge at GITEX GLOBAL](https://techxmedia.com/en/aim-intelligence-wins-supernova-challenge-at-gitex-global/) “The finale saw Korea's AIM Intelligence win the Supernova Challenge 2.0, while Chile's Chucao Technology Consultants and Ecuador's Aerialoop secured second and third places respectively.” | press | 2026-08-20 |
| s20 | [Venture Square: 에임인텔리전스, 과기정통부-구글 딥마인드 MOU 현장 참여 (names the CTO and the company's stated standardization role)](https://www.venturesquare.net/1078955) “AI 보안 전문기업 에임인텔리전스(대표 유상윤)는 과학기술정보통신부와 구글 딥마인드 간 MOU 체결식 및 전문가 간담회에 박하언 CTO가 참석해 협력 방향을 제시했다고 밝혔다.” | press | 2026-08-20 |
| s21 | [AIM Intelligence privacy policy: the legal entity name, the effective date, and the named overseas processors](https://www.aim-intelligence.com/privacy) “AIM Intelligence Inc. (hereinafter referred to as the "Company") strives to comply with the "Act on Promotion of Information and Communications Network Utilization and Information Protection" and the "Personal Information Protection Act".” | official | 2026-08-20 |
| s23 | [AIM Intelligence SUDO code repository: the published implementation of the attack framework, carrying the paper's ACL Anthology citation](https://github.com/AIM-Intelligence/SUDO) “Performs actual attack simulations based on the JSON files.” | official | 2026-08-20 |
| s22 | [Attestation probe 2026-08-20: trust. and security. subdomains and a nonsense control do not resolve, /trust, /security and /compliance return 404](https://trust.aim-intelligence.com/) | official | 2026-08-20 |

### Deep-Dive Sources

Cited from the Strategy Deep Dive section.

| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | [AIM Intelligence home page: the two products, the partner section, the deployment options, the corporate footer and the research teaser cards](https://www.aim-intelligence.com/) “AI agents are making decisions, calling APIs, and taking actions — with no one watching. AIM Intelligence attacks your AI before real attackers do, and enforces real-time guardrails to keep every agent under your control.” | official | 2026-08-20 |
| s2 | [AIM Intelligence about page: the company self-description, the by-the-numbers block, the named investor list, and the partner-programme list](https://www.aim-intelligence.com/about) “AIM Intelligence is an AI security company based in Seoul, South Korea. We develop Stinger, an automated AI red teaming platform, and Starfort, a real-time guardrail system — enabling organizations to find vulnerabilities before attackers do and protect every AI interaction in production.” | official | 2026-08-20 |
| s3 | [AIM Intelligence Stinger product page: the red-teaming platform, its stated corpus counts, its attack modes and its industry attack packs](https://www.aim-intelligence.com/stinger) “The AI red teaming platform that attacks across every modality — text, image, audio, video, and physical AI. From foundation models to autonomous agents and production applications, Stinger finds critical vulnerabilities before attackers do.” | official | 2026-08-20 |
| s4 | [AIM Intelligence Starfort product page: the guardrail engine, the policy engine, the integration modes and the deployment options](https://www.aim-intelligence.com/starfort) “Real-time guardrails that protect every AI interaction — from sensitive data leaks and malicious prompts to abnormal agent behaviors. Ultra-low latency, zero disruption, full control.” | official | 2026-08-20 |
| s5 | [AIM Intelligence research and benchmarks page: the consultancy workflow, the vendor-stated publication and assessment counts, and the named benchmark lines](https://www.aim-intelligence.com/research) “We evaluate and stress-test AI models — LLMs, VLMs, and VLAs — across safety, cultural bias, and regulatory compliance. From multi-modal threats to agentic exploits and physical AI risks, our research is published at ICLR, ICML, and ACL.” | official | 2026-08-20 |
| s6 | [AIM Intelligence press page: the company's own dated index of the coverage, awards and programme participation it claims](https://www.aim-intelligence.com/press) “AIM Intelligence in the news — from global awards and funding milestones to groundbreaking AI security research.” | official | 2026-08-20 |
| s7 | [Seoul Economic Daily: AI Security Startup Aim Intelligence Raises 10 Billion Won in Series A](https://en.sedaily.com/technology/2026/04/10/ai-security-startup-aim-intelligence-raises-10-billion-won) “Aim Intelligence, an artificial intelligence security specialist, announced Thursday that it has raised 10 billion won ($7 million) in Series A funding.” | press | 2026-08-20 |
| s8 | [thebell: AI 보안 에임인텔리전스, 100억 시리즈A 투자유치 (Korean capital-markets trade report on the Series A, sourced to the venture-capital industry)](https://www.thebell.co.kr/front/newsview.asp?key=202604090830101580101611) “에임인텔리전스는 2024년 설립된 AI 보안 전문 기업이다.” | press | 2026-08-20 |
| s9 | [ZDNet Korea: 에임인텔리전스, 100억 원 규모 시리즈A 투자 유치 성공](https://zdnet.co.kr/view/?no=20260410125429) “인공지능(AI) 보안 전문기업 에임인텔리전스(AIM Intelligence, 대표 유상윤)가 100억 원 규모의 시리즈A 투자를 유치하는 데 성공했다고 10일 밝혔다.” | press | 2026-08-20 |
| s10 | [Electronic Times: 에임인텔리전스, 100억원 규모 시리즈A 투자 유치 (hiring and SaaS build-out plans stated for the round)](https://www.etnews.com/20260410000082) “화이트해커와 LLM 엔지니어를 영입해 향상된 공격 기법과 방어 알고리즘을 갖추고, 백엔드·클라우드 엔지니어와 프론트엔드 개발자를 영입해 안정적인 서비스형소프트웨어(SaaS) 설계에 나선다.” | press | 2026-08-20 |
| s11 | [WOWTALE: AIM Intelligence Secures Pre-Series A to Advance AI Security infrastructure and Accelerate Global Expansion](https://en.wowtale.net/2025/08/09/231838/) “Its key solutions include AIM Red, an automated AI safety validation tool, AIM Guard, a real-time input/output filtering solution, and AIM Supervisor, a monitoring tool for AI agents' decision-making logic.” | press | 2026-08-20 |
| s12 | [WOWTALE: AIM Intelligence Raises $7M Series A to Scale AI Security Platform Globally](https://en.wowtale.net/2026/04/10/233852/) “The latest round brings the company's total funding to approximately $8.4M (KRW 12B).” | press | 2026-08-20 |
| s13 | [TechCrunch: The 9 top cybersecurity startups from Disrupt Startup Battlefield](https://techcrunch.com/2025/12/26/the-9-top-cybersecurity-startups-from-disrupt-startup-battlefield/) “Here is the full list of the cybersecurity Startup Battlefield 200 selectees, along with a note on why they landed in the competition.” | press | 2026-08-20 |
| s14 | [TechCrunch Startup Battlefield: the AIM Intelligence company entry for Disrupt 2025](https://techcrunch.com/startup-battlefield/company/aim-intelligence/) “AIM Intelligence AI Security and Safety for AI Agents Asia, South Korea TechCrunch Disrupt 2025” | press | 2026-08-20 |
| s15 | [ACL Anthology: sudo rm -rf agentic_security, Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics, Industry Track](https://aclanthology.org/2025.acl-industry.75/) “Anthology ID: 2025.acl-industry.75 Volume: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track) Month: July Year: 2025 Address: Vienna, Austria” | research | 2026-08-20 |
| s16 | [arXiv preprint 2503.20279: sudo rm -rf agentic_security, whose comments field records ACL 2025 Industry track acceptance](https://arxiv.org/abs/2503.20279) “[Submitted on 26 Mar 2025 ( v1 ), last revised 8 Jun 2025 (this version, v3)]” | research | 2026-08-20 |
| s17 | [Byline Network: 에임인텔리전스 AI 가드레일 전략, reported from the FISCON 2025 financial-information-protection conference](https://byline.network/2025/11/20-512/) “에임인텔리전스는 20일 여의도 콘래드 서울에서 열린 '금융정보보호 컨퍼런스 2025(FISCON 2025)'에서 '안전한 AI를 위한 전략: AI 가드레일'을 주제로 금융권 AI 시스템의 구조적 위협과 이를 통제하기 위한 기술적 대응 방안을 발표했다.” | press | 2026-08-20 |
| s18 | [Hellot: 에임인텔리전스, 2025 ACDC 핵심 프로그램 책임 (the company's role at the government-run AI hacking-defence contest and the standardization bodies it names)](https://www.hellot.net/news/article.html?no=107817) “에임인텔리전스가 국내 최대 규모 AI 보안 행사인 '2025 해킹방어대회'에서 AI 해킹방어대회 문제 출제기관으로 참여하고 AI 보안 세미나 모더레이터 역할을 맡으며 핵심 기여를 수행했다.” | press | 2026-08-20 |
| s19 | [TECHx Media: AIM Intelligence Wins Supernova Challenge at GITEX GLOBAL](https://techxmedia.com/en/aim-intelligence-wins-supernova-challenge-at-gitex-global/) “The finale saw Korea's AIM Intelligence win the Supernova Challenge 2.0, while Chile's Chucao Technology Consultants and Ecuador's Aerialoop secured second and third places respectively.” | press | 2026-08-20 |
| s20 | [Venture Square: 에임인텔리전스, 과기정통부-구글 딥마인드 MOU 현장 참여 (names the CTO and the company's stated standardization role)](https://www.venturesquare.net/1078955) “AI 보안 전문기업 에임인텔리전스(대표 유상윤)는 과학기술정보통신부와 구글 딥마인드 간 MOU 체결식 및 전문가 간담회에 박하언 CTO가 참석해 협력 방향을 제시했다고 밝혔다.” | press | 2026-08-20 |
| s21 | [AIM Intelligence privacy policy: the legal entity name, the effective date, and the named overseas processors](https://www.aim-intelligence.com/privacy) “AIM Intelligence Inc. (hereinafter referred to as the "Company") strives to comply with the "Act on Promotion of Information and Communications Network Utilization and Information Protection" and the "Personal Information Protection Act".” | official | 2026-08-20 |
| s23 | [AIM Intelligence SUDO code repository: the published implementation of the attack framework, carrying the paper's ACL Anthology citation](https://github.com/AIM-Intelligence/SUDO) “Performs actual attack simulations based on the JSON files.” | official | 2026-08-20 |
| s22 | [Attestation probe 2026-08-20: trust. and security. subdomains and a nonsense control do not resolve, /trust, /security and /compliance return 404](https://trust.aim-intelligence.com/) | official | 2026-08-20 |

## Disclaimer

This site is an experimental research aid created by Zeltser Security Corp. All its data gathering and analysis was performed autonomously without human review, and it can contain errors of fact, interpretation, and judgment that a human reviewer might catch.

The analyses are statements of opinion, not statements of fact. Machine analysis produced the scores, summaries, and matrix placements by weighing the public sources each page cites, and reasonable people can weigh the same sources differently. Where a page states a fact, it cites the public source and the date it was checked, and the statement is only as accurate as that source. Unless a profile expressly says otherwise, the analysis involves no hands-on testing and no independent validation of any company's products or services.

Nothing here is professional, security, legal, financial, investment, or purchasing advice, and nothing here is a recommendation to invest in, do business with, or avoid any company. Inclusion of a company is not an endorsement, and absence of a company is not a judgment about it. Reading this site creates no advisory or client relationship. Verify any detail you plan to act on against the vendor's current materials.

The content is provided "as is" and "as available," with all warranties disclaimed, express or implied, including merchantability, fitness for a particular purpose, accuracy, and non-infringement. No entry is warranted to be complete, current, or correct. Companies change, vendors update their claims, sources can be wrong, and automated analysis can misread them.

To the fullest extent permitted by law, the operator, Zeltser Security Corp, is not liable for any damages that arise from using this site or relying on its content, including direct, indirect, incidental, special, and consequential damages and lost profits, even if advised that such damages were possible. If you are dissatisfied with the site or disagree with these terms, your remedy is to stop using it.

Entries link to vendor pages, press coverage, and other external sites that Zeltser Security Corp does not control and is not responsible for. A link is not an affiliation with the destination or an endorsement of it. Product and company names and trademarks are the property of their owners, used here nominatively to identify the companies described. Short quotations from cited sources appear for identification and commentary.

Use, quotation, automated retrieval, and redistribution of the content are governed by the Terms of Use at cybercompanyprofiles.com/terms, which permit personal and internal business use with attribution and prohibit republication and resale.