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.
Qualifire sells developers of chatbots and AI agents small judge models that check requests and responses and can block a failing one. Qualifire also does business as Rogue Security, selling security leaders annual AWS Marketplace contracts for a platform that tests their AI agents and blocks harmful commands. CTech listed Qualifire in 2025 as founded in 2023, with $4.6 million raised and nine employees. Qualifire told CTech that checking output with another large model is slow and expensive, so it trains small judges, most for one check such as prompt injection, text that hijacks a model. A customer that switches may have to reconfigure the guardrails and policies it set up, though LiteLLM, an AI request gateway, lists rival checks from Google, CrowdStrike and Zscaler beside Qualifire.
| Description | Qualifire sells guardrails and evaluation for applications built on large language models. Small purpose-built judge models evaluate prompts, model outputs, and agent actions in real time, and a failing check can block the response before the user sees it. | [f1] |
|---|---|---|
| Founded | 2023 | [f2] |
| HQ | Tel Aviv, Israel | [f3] |
| Funding | $4.6M total | [f2] |
| Product | What it does |
|---|---|
| Qualifire Guardrails | Guardrails applied through an LLM proxy or SDK, covering prompt injection, PII, content safety, grounding, and custom policy assertions. |
| Rogue | Pre-production red teaming and evaluation for AI agents, driving multi-turn adversarial conversations over the A2A protocol and reporting what the agent did. |
| Rogue Security platform | AI agent security platform sold under the Rogue Security name, covering shadow-AI discovery, inline runtime enforcement on agent tool calls, red teaming, and MCP access control. |
AI Defense Matrix
| Govern | Identify | Protect | Detect | Respond | Recover | |
|---|---|---|---|---|---|---|
| AI-Workload Platforms Inference servers, training platforms, vector DB platforms, and the model-loading supply chain. | ||||||
| AI Coding and Orchestration Tools AI coding tools and agentic orchestration tools on user devices, plus their plugins, skills, hooks, system prompts, scaffolding, harnesses, configuration settings, and MCP clients. | ||||||
| AI-Generated Code Code produced by AI tools, AI-assisted reviews, AI-generated infrastructure-as-code and tests, and vibe-coded apps that bypass CI/CD. | ||||||
| AI Traffic Paths to external and self-hosted AI services, MCP tool channels, agent-to-agent calls, model-registry downloads, and egress to unapproved AI services. AI gateways, LLM routers, and MCP gateways steer traffic along those paths. The steering decisions and their inputs belong here too: routing policies, MCP server registries, and agent naming and discovery services. | ||||||
| AI Model Model weights, fine-tuning checkpoints, model cards, registries, an AI bill of materials (AIBOM), and the third-party LLMs your enterprise consumes. | ||||||
| Training Data Datasets used for training, fine-tuning, and continued learning. | ||||||
| Runtime AI Data User prompts, inference inputs, RAG content, vector DB content, persistent agent memory, and interaction history. | ||||||
| AI Agent Identities AI agents as non-human principals, plus credentials, keys, permission scopes, service accounts, and delegation chains across agents and tools. |
Qualifire Guardrails blocks a failing model response before it reaches users, and the Rogue Security platform discovers shadow AI and allows or blocks coding and custom agents' shell commands, tool calls, and MCP requests. These capabilities are mapped to the AI Defense Matrix. [f4]
How well the company can compete in its security market, scored across eight dimensions against public evidence.
| Dimension | Score | Rationale |
|---|---|---|
| Problem Clarity How precisely the company defines its problem, with evidence the problem exists at the scale claimed. | 3/5 | Qualifire names its buyer and the failure it addresses, teams shipping chatbots, retrieval applications and agents whose model output can break policy. Its founders gave CTech a concrete precedent, an Air Canada chatbot that a customer asked for a free ticket, after which the customer sued and won. Israel Tech Insider cites unnamed research putting hallucinations at about 20% of responses. No reviewed source measures how often deployed models break policy or what it costs, so the problem is credible and not independently quantified. [s1, s10, s31, s2] |
| Capability Depth How specific the technical capabilities are, with evidence beyond marketing claims such as docs and third-party validation. | 4/5 | The documentation names each check as an API parameter and each judge model by task, and Qualifire publishes accuracy, latency and cost figures for its models. Part of the implementation is open to inspection, because Qualifire publishes its prompt-injection model on Hugging Face behind an access agreement and the Rogue engine's source on GitHub, and LiteLLM documents the integration in its own documentation. No reviewed source has re-measured the models, and the CourtGuard preprint reports the company's own Sentinel score. [s4, s23, s5, s20, s30, s15, s21, s13] |
| Market Timing Whether the market is ready for this product, with evidence that buyers are actively seeking solutions. | 3/5 | The enabler is credible and the buyer-side evidence is indirect. Google picked the company for an AI-for-cybersecurity cohort in February 2025 and CTech reported its selection for the ScaleUp House cohort in September 2026, but both are program selections rather than records of buyers searching. No analyst category note, procurement language or regulatory driver for this product class appears in the reviewed sources, so one indirect kind of signal carries the score. [s12, s29, s10, s15] |
| Team Credibility Demonstrated domain expertise with public signals such as prior exits, publications, and industry recognition. | 3/5 | Three co-founders are named with relevant history, and CTech's 2026 profile lists the same three. Gilad Ivry told CTech he built early language-model products at Vianai, and Dror Ivry co-authored the Sentinel preprint. Ariel Dan told Israel Tech Insider that he sold his first company to Intuit in 2015 and Cloudify to Dell in early 2023. Those exits rest on his own account, with no independent reviewed source tying him to either deal, so the team sits below the corroborated tier. [s6, s10, s14, s31, s29] |
| GTM Proof Evidence of actual traction (customers, revenue signals, partnerships) beyond stated intentions. | 4/5 | LiteLLM and Portkey each document a working Qualifire integration in their own product documentation: LiteLLM's guardrail raises an HTTP 400 error on a violation, and Portkey's setup takes a Qualifire API key. Qualifire announced the Portkey partnership and an AWS Marketplace listing, and that listing names Rogue Security as the seller. Two counterparties and a marketplace make a verifiable partnership motion, multiply sourced. No reviewed outside source sizes deployments, customers or revenue, which keeps this below 5. [s15, s16, s33, s22] |
| Funding Efficiency Whether funding matches go-to-market ambition, with signs of capital-efficient growth. | 3/5 | The raise fits the motion it funds. CTech recorded $4.6 million and nine employees in September 2025, and its September 2026 profile still lists a seed-stage company of nine. Against that the company ships documented guardrails, a priced self-serve tier, a family of judge models, the Rogue engine and an agent-security platform sold through AWS Marketplace. No revenue, margin or growth figure appears in the reviewed record, so output per dollar stays unconfirmed. [s11, s29, s7, s22, s4] |
| Category Clarity Whether the company creates or fits a recognizable category that buyers can quickly place in their stack. | 3/5 | Buyers can place both products in slots that already exist. LiteLLM lists Qualifire among dozens of guardrail providers, and AWS Marketplace files the Rogue Security listing under AI Security. No analyst placement or third-party category assessment of Qualifire appears in the cited sources. [s15, s16, s22] |
| Incumbent Defensibility How vulnerable the core value proposition is to absorption as a feature by a platform vendor. | 3/5 | Qualifire says its models can be fine-tuned to a customer's domain and policies, its Sentinel paper describes private datasets targeting real-world misclassifications, Rogue runs policy and regression tests in CI/CD pipelines, and failing guardrails block responses from reaching users. Tailored checks, regression suites and inline enforcement add replacement work. LiteLLM's guardrail directory lists Bedrock Guardrails, Google Cloud Model Armor, PANW Prisma AIRS, CrowdStrike AIDR and Zscaler AI Guard beside Qualifire, so platforms could bundle the slot. [s23, s14, s30, s3, s15] |
Qualifire's argument is that offline testing cannot show how a model will behave in front of a user. Gilad Ivry told CTech that the company moves testing to runtime and decides, in the milliseconds before a user sees an answer, whether it meets the required standards. He described the product to Israel Tech Insider as a reverse firewall that checks risk going from the model out.
Two buyers appear in the record. Qualifire's documentation is addressed to developers wiring a proxy or SDK into an existing application, and on its pricing page the company offers a free plan, a $550-a-month plan and an enterprise arrangement. Under the Rogue Security name the company addresses security leaders responsible for agents their organisation uses, buys or builds, and argues that compliance teams reject cloud security proxies that route agent conversations through third-party infrastructure.
The pain is measured only in passing. Israel Tech Insider cites unnamed research putting hallucinations at about 20% of responses from advanced models. The founders' concrete case is an Air Canada chatbot that a customer asked for a free ticket, after which the customer sued and won. No research, survey or incident study in the reviewed sources counts how often deployed models break policy or what it costs when they do. [s10, s31, s2, s7, s19, s27]
Qualifire Guardrails can stop a response rather than only report on it. Qualifire documents that a failing guardrail blocks the response from reaching users, and its LiteLLM integration page records that the block setting raises an HTTP 400 on a violation while a monitor setting logs violations without blocking. Traffic reaches the checks through a proxy endpoint that stands in for the model provider's API, or through an SDK call.
The checks run on small task-specific models. Qualifire documents seven, covering prompt injection, content safety, context grounding, tool selection quality, policy enforcement, hallucination and personal data, plus a 14-billion-parameter model named Omni that handles several tasks in one call. Speed, balanced and quality modes trade latency of roughly 20, 100 and 500 milliseconds against reasoning depth.
Rogue is the pre-production engine. Qualifire describes it testing an agent over the agent-to-agent protocol with multi-turn scenarios, automating scenario creation and reporting results after each run. The source is on GitHub under a rogue-security account with a proprietary licence that allows personal and internal use and requires a licence for commercial hosting.
Under the Rogue Security name the company sells a broader agent-security platform. Its AWS Marketplace listing describes runtime enforcement inline in an agent's execution path that allows or blocks each shell command, tool call and MCP request in under five milliseconds, alongside shadow-AI discovery, red teaming and detection and response. On its site the company adds control over which AI apps and MCP tools teams may use. [s3, s21, s2, s4, s23, s9, s30, s22, s19]
Qualifire competes on cost and latency. Ariel Dan told CTech that testing model output with another large model is slow and expensive. In its benchmark table, Qualifire reports higher prompt-injection scores for Sentinel v2 than for Qwen3Guard 8B, OpenAI's GPT OSS Safeguard 20B and Meta's Llama Guard 3 8B. The table is the company's own and no reviewed independent source reproduces it, so the ranking stands as a vendor claim.
Outside researchers have used the benchmark without re-measuring the model. An arXiv preprint on a classifier called CourtGuard evaluated its own methods on the Qualifire Prompt Injection Benchmark and reported the 97.6% Sentinel figure and a 65.2% figure for ProtectAI's model, both drawn from Qualifire's own paper.
LiteLLM's guardrail directory lists Qualifire beside dozens of other entries, among them Bedrock Guardrails, Google Cloud Model Armor, PANW Prisma AIRS, CrowdStrike AIDR and Zscaler AI Guard. The listing gives Qualifire real distribution, and it also shows a buyer the alternatives it can select inside the same gateway. [s10, s23, s13, s15]
Qualifire distributes through gateways and marketplaces. LiteLLM and Portkey each document the Qualifire integration in their own documentation, and Qualifire announced its AWS Marketplace listing and its Portkey partnership on its blog in November 2025. The AWS listing names Rogue Security as the seller.
Rogue Security is a trade name of the same company. Qualifire Ltd. grants the Rogue Security service under terms of use that cover it and its affiliates doing business as Rogue Security. Qualifire Ltd. is the data controller under that site's privacy policy. CTech's September 2026 ScaleUp House profile of Rogue Security lists Ariel Dan, Dror Ivry and Gilad Ivry as founders and 2023 as the founding year. On AWS Marketplace the company prices that platform at $45,000 a year for detection and response and $115,997 for the full bundle.
On its homepage Qualifire shows four testimonials, each with a speaker's name and job title, and no reviewed outside source confirms a deployment, a customer count or revenue. Google picked the company for its 2025 AI-for-cybersecurity cohort, and CTech reported its 2026 ScaleUp House selection. [s15, s16, s33, s22, s24, s25, s29, s1, s12]
Three co-founders lead the company. On its about page Qualifire lists Ariel Dan as chief executive, Dror Ivry as chief technology officer and Gilad Ivry as chief product officer, and CTech's 2026 Rogue Security profile lists the same three in the same roles.
Gilad Ivry told CTech he worked at Vianai on products for an early generation of large language models, and Dror Ivry co-authored the Sentinel preprint. Ariel Dan told Israel Tech Insider that he sold his first company to Intuit in 2015 and, as chief executive, sold Cloudify to Dell in early 2023. Those two exits rest on his own account in the reviewed record, so they read as his claims rather than independently confirmed history.
The company is small. CTech recorded nine employees in September 2025, and its September 2026 profile lists nine again. [s6, s10, s14, s31, s11, s29, s17]
The company states a SOC 2 Type II attestation on its Rogue Security security page and says its trust center is hosted on Vanta. No audit report appears in the reviewed sources.
Qualifire's ISO 42001 page sells compliance help rather than recording a certificate. It explains what the standard asks of an AI management system, presents the product as the layer that monitors and enforces those requirements in production, and offers visitors a free certificate for signing up.
Both products give a cautious buyer control over the data path. Qualifire stores prompts or completions by default with an opt-out, encrypts logs and evaluation results, and offers on-premises or customer-cloud deployment on its enterprise plan. The Rogue Security platform offers in-VPC deployment, and the company says agent traffic then stays inside the customer's infrastructure. [s26, s18, s8, s7, s27]
| Company | Relationship | Note | Compare |
|---|---|---|---|
| Guardrails AI | competes with | Competes for the same developer choosing a guardrail layer for a shipped language-model application. | N/AWe captured the evidence for these companies under different evidence-model versions (v1 vs v2), so the totals were scored under different conditions and are not directly comparable. |
| Lakera | competes with | Competes for the same buyer looking for runtime prompt-injection defence. | |
| Guardion | competes with | Competes for the same buyer choosing guardrails for an AI application. | N/AWe captured the evidence for these companies under different evidence-model versions (v1 vs v2), so the totals were scored under different conditions and are not directly comparable. |
| Patronus AI | adjacent | Adjacent because its centre of gravity is evaluating model output rather than enforcing a decision on it. | |
| Prediction Guard | adjacent | Adjacent in the buyer's choice of controls for AI applications. |
Add analyzed competitors to compare them side by side with Qualifire.
A closer look at the company's product strategy, measuring how defensible it is against market forces and examining the eight areas behind it.
reinforce or reposition
Qualifire trains its own family of small judge models, partly on private datasets aimed at real-world misclassifications, and reports pass-or-fail checks in about twenty milliseconds. Its products sit in the path of the traffic they check, so a guardrail proxy or an inline agent control decides whether a request goes ahead. Several of its models and its prompt-injection benchmark are published on Hugging Face, which lets a security team test the detection directly. Google, CrowdStrike and Zscaler offer guardrails through the same gateway directory. A customer that switches checkers changes a gateway setting and may have to reconfigure its guardrails and policy rules.
| Dimension | Score | Rationale |
|---|---|---|
| Value Delivery Does the product sell software as the product, or judgment, trust, or accountability with software as the delivery mechanism. | 1/3 | Customers buy software their own teams configure and operate, metered by tokens on Qualifire's plans and sold as fixed annual contracts on the Rogue Security listing. Qualifire's enterprise plan includes ten AI consulting hours, a bespoke model and forward-deployed engineering. On AWS Marketplace, the company offers guided onboarding and lists a named technical account manager and quarterly security reviews for enterprise plans. The reviewed record shows no accountability for the security outcome moving to the company. |
| Switching Cost How expensive leaving is for a customer: data portability, integrations, learned workflows, network effects, regulatory data residency. | 2/3 | Leaving means real work, and the cited record does not size it. A customer accumulates configured guardrails, natural-language policy assertions and logs kept for ninety days or seven years depending on plan. Against that, the guardrails sit behind a proxy that stands in for the model provider's own interface or behind a gateway directory entry, so redirecting traffic to another checker is a configuration change. |
| Compliance Moat Whether certifications, liability acceptance, or audit trails block an easy replacement. | 1/3 | On its Rogue Security security page the company states a SOC 2 Type II attestation and a trust center hosted on Vanta, and no audit report appears in the reviewed sources. A SOC 2 report is an ordinary enterprise credential a funded competitor can obtain, and Qualifire's ISO 42001 page sells help with the customer's own compliance rather than recording a certificate the company holds. Nothing in the reviewed record blocks a replacement on compliance grounds. |
| Problem Complexity Whether the product requires ML, optimization, real-time systems, or years of specialized expertise. | 3/3 | The work is machine learning under a real-time budget. Qualifire trains and serves task-specific judge models and a 14-billion-parameter multi-task model, and documents pass-or-fail checks near twenty milliseconds, while its AWS listing describes micro language models deciding on agent actions in under five milliseconds. Holding accuracy at those latencies and costs takes specialised engineering rather than assembly. |
| Buyer Profile Whether buyers are SMB operators, mid-market IT teams, or regulated enterprises and governments with procurement gates. | 2/3 | The company addresses two buyers. Qualifire's guardrails start with a free self-serve account and a $550-a-month plan for developers. The Rogue Security platform addresses security leaders with fixed annual contracts of $45,000 and $115,997 on AWS Marketplace, in-VPC deployment and a stated SOC 2 attestation. The reviewed record names no regulated or government buyer and no channel aimed at one, so the blend sits at the mid-market level. |
| Layer Whether the product is an end-user application, a platform with application features, or infrastructure other applications depend on. | 3/3 | Both products sit between the customer's applications or agents and the systems they call, for the traffic routed through them. Qualifire documents a proxy that stands in for the model provider's API and guardrails that block a failing response before it reaches users, with the LiteLLM integration returning an HTTP 400 on a violation. In its Rogue Security listing, the company places its runtime product inline in the agent's execution path. The dependence covers only traffic customers route through those components, and log or monitor modes let requests proceed. |
| Proprietary Data, Content, or IP Whether the product accumulates datasets, content licenses, or IP that a rival cannot recreate from scratch. | 2/3 | The record evidences a vendor-held training asset without naming or sizing it. The Sentinel preprint by the company's chief technology officer says the model was fine-tuned on open-source and private collections, with private datasets aimed at real-world misclassifications. Gilad Ivry told an interviewer the company iterates on customer feedback about labelling. Several models and the benchmark are public on Hugging Face. Qualifire used private training material for Sentinel. |
Qualifire sells to two segments under two names. The guardrails product is adopted by the developer who owns a shipped AI feature and wires a proxy or SDK into it, starting from a free account. The Rogue Security platform is pitched to security leaders, with its site organised around AI an organisation uses, buys and builds.
The tiers show the split. Qualifire's Basic plan costs nothing and caps tokens, protection rules and log retention, and its Pro plan costs $550 a month with a billion tokens and ninety days of logs. Its enterprise plan is quoted and adds customer-cloud or on-premises hosting, seven years of logs and consulting hours. The Rogue Security platform sells on twelve-month contracts of $45,000 and $115,997 on AWS Marketplace.
No reviewed source names a buyer in a regulated industry or in government. The homepage testimonials give names and job titles, and the reviewed sources do not say which plan any customer bought, so the buyers of the enterprise tier are unconfirmed.
The technical bet is that the checker should be small. Ariel Dan told CTech the company uses small language models or classic machine learning rather than a large model grading output. Qualifire documents seven task-specific judges and a 14-billion-parameter multi-task model, with speed, balanced and quality modes trading latency against reasoning depth.
Enforcement is a decision rather than a report. Qualifire documents that a failing guardrail blocks the response, and its LiteLLM integration returns an HTTP 400 on a violation. For agents, the company's AWS listing describes micro language models that allow or block each shell command, tool call and MCP request in under five milliseconds. CTech's 2026 profile describes an intent classification engine called RISC that runs inside the customer's network.
The CourtGuard researchers evaluated their classifier on Qualifire's benchmark and cited Qualifire's reported 97.6% result for Sentinel, its prompt-injection detector. Qualifire also publishes its own comparison placing Sentinel v2 above Qwen3Guard, GPT OSS Safeguard and Llama Guard 3. No reviewed source has measured the models independently.
Distribution runs through other companies' gateways and marketplaces. LiteLLM and Portkey each carry a Qualifire page in their own documentation, and Portkey's page describes more than twenty guardrail checks. Qualifire announced its AWS Marketplace listing and its Portkey partnership in November 2025 and its LiteLLM availability in January 2026.
The newer motion sells under the Rogue Security name. Qualifire Ltd. grants the Rogue Security service under terms covering it and its affiliates doing business as Rogue Security. Qualifire Ltd. is the data controller under the rogue.security privacy policy. On AWS Marketplace, the company sells as Rogue Security. CTech's September 2026 ScaleUp House profile describes Rogue Security as a seed-stage company starting with coding agents.
References are self-presented. On its homepage Qualifire shows four testimonials with names and job titles, and no reviewed outside source confirms a deployment. Google's 2025 AI-for-cybersecurity cohort and the 2026 ScaleUp House cohort are the third-party selections on record.
Qualifire publishes guardrail prices. A free plan includes three million tokens a month. A Basic plan at no cost allows 300,000 tokens, three protection rules and seven days of logs. A Pro plan at $550 a month includes a billion tokens, unlimited rules and ninety days of logs, with extra tokens billed per million. An enterprise plan is quoted and adds customer-cloud or on-premises hosting and seven years of logs.
The Rogue Security platform prices the other way. On AWS Marketplace the company offers twelve-month contracts of $45,000 for detection and response and $115,997 for the full bundle with an engineering deployment package, and describes them as fixed annual pricing with unlimited usage, never per-token.
Two details shape a buyer's choice of plan. Personal-data masking appears on the Pro plan, and the log window runs to ninety days below the enterprise tier, and the quoted enterprise plan lists seven years of log retention.
Qualifire delivers the guardrails as a hosted service the customer points traffic at, through a proxy that stands in for the model provider's API or through an SDK. On its homepage Qualifire offers the customer's own cloud, an on-premises install or the hosted service.
An inline check costs latency, and Qualifire puts numbers on it. Its documentation records roughly twenty milliseconds for a pass-or-fail check, about one hundred with reasoning and about five hundred with larger models. Those figures are the company's own.
The Rogue Security platform runs in the customer's VPC, on premises or as SaaS. The company offers an SDK and CI/CD integration, and on AWS Marketplace promises round-the-clock support with a four-hour response target for critical issues and guided onboarding with deployment assistance. On the guardrails side, Qualifire stores prompts or completions by default with an opt-out, which a team sending confidential material would change first.
The company states a SOC 2 Type II attestation on its Rogue Security security page and says its trust center is hosted on Vanta. No audit report appears in the reviewed sources.
Qualifire markets ISO 42001 compliance assistance. It explains what the standard asks of an AI management system and presents the product as the layer that monitors and enforces those requirements, and it offers visitors a free certificate for signing up.
A cautious buyer has deployment control and inspectable models to lean on. The enterprise plan offers customer-cloud or on-premises hosting, and the prompt-injection model is published on Hugging Face behind an access agreement, so a security team can test the detection itself.
Qualifire's guardrails are a component inside other companies' control points. The product offers a proxy that stands in for a provider's own endpoint and appears in the guardrail directories of two gateway projects.
Those directories double as a list of alternatives. LiteLLM's carries Qualifire beside dozens of entries, among them guardrails from Bedrock, Google Cloud Model Armor, PANW Prisma AIRS, CrowdStrike and Zscaler.
The Rogue engine and the Rogue Security platform are also company products. The engine's source is on GitHub under a licence free for internal use, and the company says on AWS Marketplace that the platform has 220 or more integrations across SDK, gateway, sidecar, browser extension and IDE hooks. That number describes the connector catalogue and says nothing about how many customers use those connectors.
Three co-founders run the company. Ariel Dan is chief executive, Dror Ivry chief technology officer and Gilad Ivry chief product officer, and CTech's 2026 Rogue Security profile lists the same three.
Gilad Ivry told CTech he worked at Vianai on early language-model products, and Dror Ivry co-authored the Sentinel preprint. Ariel Dan told Israel Tech Insider he sold a company to Intuit in 2015 and Cloudify to Dell in early 2023, which rests on his own account in the reviewed record.
The company was small when last counted. CTech recorded nine employees and $4.6 million raised in September 2025, and its 2026 profile lists nine employees and a seed stage.
| Id | Source | Tier | Accessed |
|---|---|---|---|
| f1 | Qualifire: homepage | official | 2026-10-23 |
| f2 | CTech: Qualifire, an AI governance platform ensuring businesses receive accurate responses from chatbots | press | 2026-10-23 |
| f3 | Google: 16 startups using AI to make our world safer | press | 2026-10-23 |
| f4 | AWS Marketplace: Rogue Security runtime security listing | official | 2026-10-23 |
| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | Qualifire: homepage “Continuous evaluation, real time guardrails and pre production agentic testing. Made for agents, RAG and chatbots.” | official | 2026-10-23 |
| s2 | Qualifire docs: introduction “Contextual, real-time guardrails that protect your application from safety, security, and policy violations while taking context into account.” | official | 2026-10-23 |
| s3 | Qualifire docs: Guardrails “Guardrails that fail will block the response from reaching your users. Make sure to test your guardrail configuration thoroughly before deploying to production.” | official | 2026-10-23 |
| s4 | Qualifire docs: Evaluations and the judge models “Qualifire employs a suite of fine-tuned, state-of-the-art Small Language Models (SLMs), each specialized for a specific evaluation task.” | official | 2026-10-23 |
| s5 | Qualifire docs: Evaluate endpoint API reference “Evaluates given input, output, or messages using Qualifire's detectors. Supports checks for hallucinations, grounding, PII, prompt injections, content moderation, policy assertions, topic scoping, and tool use quality.” | official | 2026-10-23 |
| s6 | Qualifire: about page “Qualifire provides real-time guardrails for LLM applications by preventing hallucinations, moderating content, and enforcing policy guidelines.” | official | 2026-10-23 |
| s7 | Qualifire: pricing page “Pro $550 / month” | official | 2026-10-23 |
| s8 | Qualifire: ISO 42001 page “Qualifire monitors and enforces AI compliance in production – in line with ISO/IEC 42001:2024.” | official | 2026-10-23 |
| s9 | Qualifire: Rogue product page “End to End Testing Framework for Agentic Systems” | official | 2026-10-23 |
| s10 | CTech: The Israeli startup tackling AI unpredictable behavior “The moment the model delivers an answer, milliseconds before the user sees it, we analyze it and decide if the response meets the required standards. If we detect a violation, it is blocked, preventing the customer from experiencing it” | press | 2026-10-23 |
| s11 | CTech: Qualifire, an AI governance platform ensuring businesses receive accurate responses from chatbots “Founded: 2023” | press | 2026-10-23 |
| s12 | Google: 16 startups using AI to make our world safer “Qualifire (Tel Aviv, Israel): Qualifire evaluates LLM-based applications in real time, ensuring safety, reliability and compliance.” | press | 2026-10-23 |
| s13 | arXiv preprint: CourtGuard, a local multiagent prompt injection classifier “Qualifire’s Sentinel model, which was self proclaimed as state-of-the-art in its release in June, 2025, scores 97.6% on the Qualifire Prompt Injection Benchmark—much higher than all instances of the Direct Detector or CourtGuard ( Ivry & Nahum, 2025 )” | research | 2026-10-23 |
| s14 | arXiv preprint: Sentinel, a model to protect against prompt injections “By leveraging ModernBERT's advanced features and fine-tuning on an extensive and diverse dataset comprising a few open-source and private collections, Sentinel achieves state-of-the-art performance.” | research | 2026-10-23 |
| s15 | LiteLLM docs: Qualifire guardrail provider page “Qualifire supports the following evaluation checks:” | research | 2026-10-23 |
| s16 | Portkey docs: Qualifire guardrails page on a partner gateway “Qualifire provides comprehensive AI reliability and quality checks including content moderation, hallucination detection, and policy compliance.” | official | 2026-10-23 |
| s17 | Hugging Face: the qualifire organization page “qualifire (Rogue Security)” | official | 2026-10-23 |
| s18 | Qualifire : trust probe 2026-10-23 by capture-extract, /trust and /security 404, trust., security. and control subdomains unresolved | official | 2026-10-23 |
| s19 | Rogue Security: homepage “Real-time protection and policy enforcement for every agent, everywhere.” | official | 2026-10-23 |
| s20 | Hugging Face: the Sentinel prompt-injection model card “The model supports secure LLM deployments by acting as a gatekeeper to filter potentially adversarial user inputs.” | official | 2026-10-23 |
| s21 | Qualifire docs: LiteLLM guardrails integration “on_flagged: "block" raises an HTTP 400 exception when violations are detected” | official | 2026-10-23 |
| s22 | AWS Marketplace: Rogue Security runtime security listing “Sold by Rogue Security” | official | 2026-10-23 |
| s23 | Qualifire docs: SLM judges and the vendor benchmark comparison “Benchmark comparison (Prompt Injection): Model Creator Avg F1 Latency Cost/1M tokens Sentinel v2 Qualifire 0.957 ~0.038s $0.005” | official | 2026-10-23 |
| s24 | Rogue Security: terms of use “ROGUE SECURITY – TERMS OF USE” | official | 2026-10-23 |
| s25 | Rogue Security: privacy policy “At Qualifire Ltd., its subsidiaries (if any) or affiliated companies ("Qualifire" or "We") as the controller of the data, we recognize that privacy is important.” | official | 2026-10-23 |
| s26 | Rogue Security: security and trust page “Our Trust Center is hosted on Vanta and provides detailed information about our security posture, SOC 2 compliance, and data handling practices.” | official | 2026-10-23 |
| s27 | Rogue Security: AI AppSec product page “Red team your agents. Deploy guardrails. Ship secure.” | official | 2026-10-23 |
| s28 | Rogue Security: AIDR product page “Real-time detection and response for every AI agent interaction.” | official | 2026-10-23 |
| s29 | CTech: Rogue Security catches malicious AI coding agents in milliseconds “Rogue Security is one of the companies selected for the inaugural ScaleUp House cohort, joining the year-round program as it enters its next stage of international growth.” | press | 2026-10-23 |
| s30 | GitHub: the Rogue repository “GitHub - rogue-security/rogue: AI Agent Evaluator & Red Team Platform · GitHub” | official | 2026-10-23 |
| s31 | Israel Tech Insider: The Race to Stop AI Hallucinations, by Amir Mizroch “Co-founded my first company in 2012. Sold it to Intuit 2015. Joined Cloudify as a CEO. And we sold that company to Dell about two years ago in early 2023 and Qualifire.” | press | 2026-10-23 |
| s33 | Qualifire: blog index “We're excited to announce that Qualifire is partnering with Portkey to bring production-ready guardrails to the Portkey LLM Gateway.” | official | 2026-10-23 |
| s34 | Qualifire: terms of use “contains the TOU upon which Qualifire Ltd., and its affiliates (“ Qualifire ”), grants to you” | official | 2026-10-23 |
| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | Qualifire: homepage “Continuous evaluation, real time guardrails and pre production agentic testing. Made for agents, RAG and chatbots.” | official | 2026-10-23 |
| s2 | Qualifire docs: introduction “Contextual, real-time guardrails that protect your application from safety, security, and policy violations while taking context into account.” | official | 2026-10-23 |
| s3 | Qualifire docs: Guardrails “Guardrails that fail will block the response from reaching your users. Make sure to test your guardrail configuration thoroughly before deploying to production.” | official | 2026-10-23 |
| s4 | Qualifire docs: Evaluations and the judge models “Qualifire employs a suite of fine-tuned, state-of-the-art Small Language Models (SLMs), each specialized for a specific evaluation task.” | official | 2026-10-23 |
| s5 | Qualifire docs: Evaluate endpoint API reference “Evaluates given input, output, or messages using Qualifire's detectors. Supports checks for hallucinations, grounding, PII, prompt injections, content moderation, policy assertions, topic scoping, and tool use quality.” | official | 2026-10-23 |
| s6 | Qualifire: about page “Qualifire provides real-time guardrails for LLM applications by preventing hallucinations, moderating content, and enforcing policy guidelines.” | official | 2026-10-23 |
| s7 | Qualifire: pricing page “Pro $550 / month” | official | 2026-10-23 |
| s8 | Qualifire: ISO 42001 page “Qualifire monitors and enforces AI compliance in production – in line with ISO/IEC 42001:2024.” | official | 2026-10-23 |
| s9 | Qualifire: Rogue product page “End to End Testing Framework for Agentic Systems” | official | 2026-10-23 |
| s10 | CTech: The Israeli startup tackling AI unpredictable behavior “The moment the model delivers an answer, milliseconds before the user sees it, we analyze it and decide if the response meets the required standards. If we detect a violation, it is blocked, preventing the customer from experiencing it” | press | 2026-10-23 |
| s11 | CTech: Qualifire, an AI governance platform ensuring businesses receive accurate responses from chatbots “Founded: 2023” | press | 2026-10-23 |
| s12 | Google: 16 startups using AI to make our world safer “Qualifire (Tel Aviv, Israel): Qualifire evaluates LLM-based applications in real time, ensuring safety, reliability and compliance.” | press | 2026-10-23 |
| s13 | arXiv preprint: CourtGuard, a local multiagent prompt injection classifier “Qualifire’s Sentinel model, which was self proclaimed as state-of-the-art in its release in June, 2025, scores 97.6% on the Qualifire Prompt Injection Benchmark—much higher than all instances of the Direct Detector or CourtGuard ( Ivry & Nahum, 2025 )” | research | 2026-10-23 |
| s14 | arXiv preprint: Sentinel, a model to protect against prompt injections “By leveraging ModernBERT's advanced features and fine-tuning on an extensive and diverse dataset comprising a few open-source and private collections, Sentinel achieves state-of-the-art performance.” | research | 2026-10-23 |
| s15 | LiteLLM docs: Qualifire guardrail provider page “Qualifire supports the following evaluation checks:” | research | 2026-10-23 |
| s16 | Portkey docs: Qualifire guardrails page on a partner gateway “Qualifire provides comprehensive AI reliability and quality checks including content moderation, hallucination detection, and policy compliance.” | official | 2026-10-23 |
| s17 | Hugging Face: the qualifire organization page “qualifire (Rogue Security)” | official | 2026-10-23 |
| s18 | Qualifire and Rogue Security: trust probe 2026-10-23 by capture-extract, /trust and /security 404, trust., security. and control subdomains unresolved | official | 2026-10-23 |
| s19 | Rogue Security: homepage “Real-time protection and policy enforcement for every agent, everywhere.” | official | 2026-10-23 |
| s20 | Hugging Face: the Sentinel prompt-injection model card “The model supports secure LLM deployments by acting as a gatekeeper to filter potentially adversarial user inputs.” | official | 2026-10-23 |
| s21 | Qualifire docs: LiteLLM guardrails integration “on_flagged: "block" raises an HTTP 400 exception when violations are detected” | official | 2026-10-23 |
| s22 | AWS Marketplace: Rogue Security runtime security listing “Sold by Rogue Security” | official | 2026-10-23 |
| s23 | Qualifire docs: SLM judges and the vendor benchmark comparison “Benchmark comparison (Prompt Injection): Model Creator Avg F1 Latency Cost/1M tokens Sentinel v2 Qualifire 0.957 ~0.038s $0.005” | official | 2026-10-23 |
| s24 | Rogue Security: terms of use “ROGUE SECURITY – TERMS OF USE” | official | 2026-10-23 |
| s25 | Rogue Security: privacy policy “At Qualifire Ltd., its subsidiaries (if any) or affiliated companies ("Qualifire" or "We") as the controller of the data, we recognize that privacy is important.” | official | 2026-10-23 |
| s26 | Rogue Security: security and trust page “Our Trust Center is hosted on Vanta and provides detailed information about our security posture, SOC 2 compliance, and data handling practices.” | official | 2026-10-23 |
| s27 | Rogue Security: AI AppSec product page “Red team your agents. Deploy guardrails. Ship secure.” | official | 2026-10-23 |
| s28 | Rogue Security: AIDR product page “Real-time detection and response for every AI agent interaction.” | official | 2026-10-23 |
| s29 | CTech: Rogue Security catches malicious AI coding agents in milliseconds “Rogue Security is one of the companies selected for the inaugural ScaleUp House cohort, joining the year-round program as it enters its next stage of international growth.” | press | 2026-10-23 |
| s30 | GitHub: the Rogue repository “GitHub - rogue-security/rogue: AI Agent Evaluator & Red Team Platform · GitHub” | official | 2026-10-23 |
| s31 | Israel Tech Insider: The Race to Stop AI Hallucinations, by Amir Mizroch “Co-founded my first company in 2012. Sold it to Intuit 2015. Joined Cloudify as a CEO. And we sold that company to Dell about two years ago in early 2023 and Qualifire.” | press | 2026-10-23 |
| s33 | Qualifire: blog index “We're excited to announce that Qualifire is partnering with Portkey to bring production-ready guardrails to the Portkey LLM Gateway.” | official | 2026-10-23 |
| s34 | Qualifire: terms of use “contains the TOU upon which Qualifire Ltd., and its affiliates (“ Qualifire ”), grants to you” | official | 2026-10-23 |
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.