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 analysis is scoped to OpenAI Guardrails.
OpenAI Guardrails is a free, MIT-licensed library that checks the inputs and outputs of AI applications and agents developers build on OpenAI's APIs. OpenAI released it in October 2025 and labels its Python package a preview. Its checks flag jailbreak attempts, prompt injection in tool calls, and personal data. Checks that ask an OpenAI model to judge content call OpenAI's paid APIs at developers' cost. OpenAI itself closed a $122 billion round at an $852 billion valuation. HiddenLayer researchers reported bypassing the jailbreak and prompt injection checks by prompt-injecting the judging model and the base model at the same time. Developers operate the checks inside their own applications and own the results.
| Description | OpenAI Guardrails adds configurable safety checks to LLM applications, detecting jailbreak attempts on model calls, masking personally identifiable information in text, and moderating content that classifiers flag. | [f1] |
|---|---|---|
| Latest funding | $122B round at $852B post-money valuation (closed March 2026) | [f2] |
| Deployment | SaaS, Self-hosted | [f3] |
| Product | What it does |
|---|---|
| OpenAI Guardrails | OpenAI Guardrails: Safety framework that validates LLM app inputs and outputs with configurable checks, plus open-weight gpt-oss-safeguard policy classifiers. |
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. |
OpenAI Guardrails validates LLM app inputs and outputs with configurable checks, plus open-weight gpt-oss-safeguard policy classifiers. It is 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 | OpenAI names developers adding safety checks to LLM applications and agents as the buyer, and it targets jailbreaks, prompt injection in tool calls, and exposure of personal data. HiddenLayer, Zenity Labs, and a CISA alert treat attacks on AI systems as real, but none of the cited sources quantifies what the problem costs buyers, so the problem is clear while its cost is not independently measured. [s1, s7, s13, s14, s17] |
| Capability Depth How specific the technical capabilities are, with evidence beyond marketing claims such as docs and third-party validation. | 4/5 | OpenAI documents each check in detail, including a Jailbreak benchmark built from public datasets and synthetic conversations, and the MIT-licensed Python code is open to inspection. HiddenLayer tested the framework and Zenity Labs tested the guardrails in OpenAI’s AgentKit, so outside scrutiny of the capability is multiply evidenced, even though both teams reported bypasses of specific checks. [s1, s3, s7, s8, s13, s14] |
| Market Timing Whether the market is ready for this product, with evidence that buyers are actively seeking solutions. | 3/5 | The enabler is AI agents that call tools, which the Prompt Injection Detection check exists to police, and Zenity Labs published its test of OpenAI’s AgentKit guardrails on October 10, 2025. That is security-research interest rather than buyer-side demand, and the one policy signal in the cited sources is a CISA alert of April 15, 2024, so timing is plausible while demand evidence is indirect. [s14, s13, s17, s7] |
| Team Credibility Demonstrated domain expertise with public signals such as prior exits, publications, and industry recognition. | 3/5 | No captured source mentions an individual who leads the Guardrails framework, and its release notes list contributors by GitHub handle. OpenAI’s company-level safety work is verifiable, including gpt-oss-safeguard built with Discord, SafetyKit, and ROOST, but no captured source links that work to the Guardrails maintainers, so the cited sources support relevant experience rather than prior builds by the line’s own team. [s6, s15, s16, s18] |
| GTM Proof Evidence of actual traction (customers, revenue signals, partnerships) beyond stated intentions. | 3/5 | No captured source mentions a customer of OpenAI Guardrails, and OpenAI distributes it to developers through PyPI, GitHub, and the Agents SDK integration, which on its own would score as thin traction. The score applies the allowed indirect-signal adjustment for OpenAI’s platform distribution. [s5, s1] |
| Funding Efficiency Whether funding matches go-to-market ambition, with signs of capital-efficient growth. | 3/5 | This is a product-line score, so OpenAI’s own funding is not credited to Guardrails, and the cited sources show no separate budget or line economics. PyPI’s release history records releases from October 6, 2025 through September 10, 2026, three of them from July 2026 onward, so shipping is visible while efficiency stays unconfirmed. [s5, s6] |
| Category Clarity Whether the company creates or fits a recognizable category that buyers can quickly place in their stack. | 3/5 | OpenAI Guardrails fits the recognizable class of LLM guardrails, and HiddenLayer and Zenity Labs each evaluated OpenAI’s guardrails as that kind of product. That placement comes from security researchers rather than buyers or analysts, and no analyst category note appears in the cited sources, so the category is recognizable but not multiply established. [s13, s14, s2] |
| Incumbent Defensibility How vulnerable the core value proposition is to absorption as a feature by a platform vendor. | 2/5 | This scores the Guardrails line’s own moat rather than OpenAI’s size. OpenAI built LLM-based detectors and benchmarked the Jailbreak check, which is more than a trivial checkbox, but the code is MIT licensed, so a platform already serving the same developers could adopt or reimplement comparable checks in an ordinary release. [s8, s1] |
OpenAI Guardrails addresses the risk that LLM applications and agents act on hostile or sensitive input. According to its repository, it adds configurable safety and compliance checks that validate inputs and outputs, and its Prompt Injection Detection check confirms that an agent’s tool calls match the user’s goal before they run.
Outside sources treat the problem as real. HiddenLayer describes the framework as pipelines that detect, filter, or block harmful inputs, outputs, and tool calls. A CISA alert from April 2024 lists methodologies and controls against malicious activity targeting AI systems. [s1, s7, s13, s17]
OpenAI Guardrails is a pipeline of configurable checks that wraps the OpenAI client. The repository README lists twelve guardrails, among them Secret Keys, Contains PII, Jailbreak, Prompt Injection Detection, and Hallucination Detection. The Python and TypeScript documentation describe validation across input, output, and pre-flight stages.
Developers build a configuration in the hosted wizard or define it directly, then use a drop-in client or the GuardrailAgent integration with the Agents SDK. The wizard page lists Jailbreak as LLM-based and Prompt Injection Detection as Experimental. OpenAI’s documentation includes a benchmark for the Jailbreak check, and an evaluation tool measures precision, recall, and F1 against labeled data. [s1, s2, s3, s4, s7, s8, s9]
OpenAI Guardrails reaches developers as a free, MIT-licensed library built around OpenAI’s own client and Agents SDK. Several of its checks call an OpenAI model, and according to the README, Guardrails calls paid OpenAI APIs, so developers who adopt those checks pay for OpenAI API usage.
The outside evaluations in the reviewed sources come from security research, where HiddenLayer and Zenity Labs tested OpenAI’s guardrails and reported ways around specific checks. [s1, s8, s13, s14]
OpenAI distributes Guardrails through PyPI, GitHub, and a hosted configuration wizard, and no captured source mentions a customer. PyPI’s release history records releases from October 6, 2025 through September 10, 2026, and its classifiers list developers as the intended audience.
After no release between December 15, 2025 and July 21, 2026, OpenAI shipped three releases that include Python 3.14 support and fixes to secret detection, PII masking, and prompt injection detection. On June 3, 2026, OpenAI told developers it is deprecating Agent Builder, which offered built-in guardrail nodes, and it points them to the Agents SDK, which the Guardrails library integrates with. [s5, s6, s1, s10, s11]
OpenAI publishes the Guardrails framework, and no captured source mentions the people who lead it. The GitHub release notes list contributors by handle without describing their backgrounds.
OpenAI’s safety work around the framework is documented at the company level. OpenAI developed the gpt-oss-safeguard models with Discord, SafetyKit, and ROOST, and ROOST describes that release as one of the core components of OpenAI’s safety stack. The US and UK AI Safety Institutes also shared pre-deployment test findings on an OpenAI model with the company. No captured source links these efforts to the team that maintains Guardrails, so they count as OpenAI’s safety record rather than that team’s. [s6, s15, s16, s18]
OpenAI publishes the Guardrails code openly and makes developers responsible for operating it safely. The code is MIT licensed, and users of the wizard agree to the OpenAI Services Agreement, which governs the use of Guardrails. Under the README disclaimers, developers are responsible for safeguards against storing sensitive or prohibited content.
The Python repository and PyPI package carry a Preview label, while the Python and TypeScript documentation describe Guardrails as production ready, so adopters see conflicting maturity labels. In its agent-safety guide, OpenAI calls built-in guardrail nodes an effective first wave of protection that is not foolproof. HiddenLayer advises organizations to treat LLM-based detection as a supplement to other defenses. [s1, s2, s3, s5, s11, s13]
| Company | Relationship | Note | Compare |
|---|---|---|---|
| Lakera | competes with | Competes for the same developer teams adding prompt-injection checks. | |
| Meta LlamaFirewall | competes with | Competes for the same application developers. | |
| Prompt Security | competes with | Competes for teams that want runtime checks on LLM applications. | |
| NVIDIA NeMo Guardrails | competes with | An alternative for developers choosing a guardrail library. |
Add analyzed competitors to compare them side by side with OpenAI.
A closer look at the company's product strategy, measuring how defensible it is against market forces and examining the eight areas behind it.
pivot urgently
OpenAI gives Guardrails away as code under the MIT License that developers run inside their own applications, and it publishes gpt-oss-safeguard models that classify text against developer-provided safety policies under Apache 2.0. Leaving the library would likely mean replacing the client wrapper and adapting configurations. Developers accept the OpenAI Services Agreement to use the wizard and are responsible for compliance under the library’s disclaimers.
| 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 | OpenAI delivers Guardrails as software developers configure and run in their own applications, so the customer’s team operates the checks and owns the outcomes, which is the software-product level. |
| Switching Cost How expensive leaving is for a customer: data portability, integrations, learned workflows, network effects, regulatory data residency. | 2/3 | Leaving Guardrails would likely mean replacing the drop-in client or GuardrailAgent wrapper and adapting the exported pipeline configuration and any evaluation datasets whose labels map guardrail names, which is integration and workflow friction. The cited sources do not size that migration, and they document no network effect or data residency tie that would make leaving expensive. |
| Compliance Moat Whether certifications, liability acceptance, or audit trails block an easy replacement. | 1/3 | No captured source names a certification or regulatory mandate for Guardrails, and under its README disclaimers developers must ensure their own systems comply with applicable laws. |
| Problem Complexity Whether the product requires ML, optimization, real-time systems, or years of specialized expertise. | 2/3 | Guardrails wraps the OpenAI client and assembles validation pipelines that ask existing OpenAI models to judge conversations and tool calls, with an evaluation tool that reports precision, recall and F1. That is moderate algorithmic depth. The cited record does not show that rebuilding the framework requires years of specialized expertise, and OpenAI’s own model development is not credited to the line. |
| Buyer Profile Whether buyers are SMB operators, mid-market IT teams, or regulated enterprises and governments with procurement gates. | 2/3 | The captured sources describe developers building LLM applications on OpenAI’s API, including Azure clients, who agree to the OpenAI Services Agreement to use the wizard, and no captured source names a regulated enterprise or government buyer. Adopter size is not evidenced, so the score stays at the middle level rather than reaching procurement-gated buyers. |
| Layer Whether the product is an end-user application, a platform with application features, or infrastructure other applications depend on. | 2/3 | Guardrails wraps the OpenAI client inline on every API call and adds a configuration wizard and an evaluation tool, which makes it a platform with application features. An application can run on the plain OpenAI client without it. |
| Proprietary Data, Content, or IP Whether the product accumulates datasets, content licenses, or IP that a rival cannot recreate from scratch. | 1/3 | No captured source names a retained non-public dataset, content license, or patent behind Guardrails. OpenAI built its Jailbreak evaluation benchmark from public datasets plus synthetic conversations to measure the check, the code is MIT licensed, and the gpt-oss-safeguard weights are published under Apache 2.0, so no captured source evidences a retained asset the product accumulates. |
OpenAI Guardrails targets developers building LLM applications and agents on OpenAI’s APIs. PyPI lists developers as the intended audience, the library wraps the OpenAI Python client, and the wizard includes starting configurations for customer support, internal research, and coding assistants.
The captured sources describe adopters by integration path, since developers can use the drop-in client or the Agents SDK integration.
OpenAI Guardrails combines rule-based checks with checks that ask an OpenAI model to judge content. The README lists twelve guardrails, and the wizard lists Jailbreak as LLM-based and Mask PII as hybrid. The Jailbreak check reads conversation history for multi-turn escalation, and Prompt Injection Detection checks tool calls before they run.
OpenAI documents the model-based work and publishes part of it. OpenAI built its Jailbreak benchmark from public datasets plus synthetic benign conversations, and the related gpt-oss-safeguard models apply a developer’s policy at inference time with a chain of thought developers can read.
OpenAI distributes Guardrails as free open-source software through developer channels. Developers install it from PyPI or GitHub, configure it in a hosted wizard, and connect it to OpenAI’s client or the Agents SDK, and no captured source mentions a customer.
PyPI’s release history records no release between December 15, 2025 and July 21, 2026, and three releases followed through September 10, 2026. OpenAI is also deprecating Agent Builder, which offered guardrail nodes, and it directs those developers to the Agents SDK or ChatGPT Workspace Agents.
OpenAI charges no license fee for Guardrails, and adopters pay for the OpenAI API calls its checks make. The code is MIT licensed, and according to the README, Guardrails calls paid OpenAI APIs and developers are responsible for the charges. The gpt-oss-safeguard models ship under Apache 2.0.
Cost depends on which checks a developer enables and how they are configured. HiddenLayer reports that OpenAI’s recommended default for the Jailbreak check is the gpt-4.1-mini model, and OpenAI’s documentation advises keeping reasoning output disabled in production to minimize costs and latency.
Developers run Guardrails inside their own applications, configuring it in the hosted wizard or defining pipeline configurations directly. The Python documentation describes a drop-in replacement for OpenAI clients whose checks run automatically on every API call. The Contains PII check needs a spaCy model installed during build or deployment.
OpenAI ships the library through tagged releases, and its September 10, 2026 release included fixes for several detection defects plus automated release tooling and CodeQL scanning. The reviewed repository and documentation describe no managed service that runs the checks for a customer, so the adopting team operates them.
Developers can inspect OpenAI Guardrails, and OpenAI documents its limits. The MIT-licensed code is on GitHub, the gpt-oss-safeguard models produce a chain of thought developers can read, and in its agent-safety guide OpenAI says built-in guardrail nodes are not foolproof.
Independent researchers found weaknesses in specific checks. HiddenLayer focused on the Jailbreak and Prompt Injection Detection checks and bypassed both, and Zenity Labs reported that a simple change of pattern bypassed the PII guardrail in OpenAI’s AgentKit. Under the README disclaimers, developers are responsible for safeguarding sensitive content, and users of the wizard agree to the OpenAI Services Agreement, which governs the use of Guardrails.
OpenAI Guardrails is built for OpenAI’s developer platform. It wraps the OpenAI client, integrates with the Agents SDK through GuardrailAgent, and calls paid OpenAI APIs for its model-based checks.
OpenAI is shutting down Agent Builder on November 30, 2026 and directing developers to the Agents SDK or ChatGPT Workspace Agents, and the Guardrails README documents an integration with the Agents SDK.
OpenAI maintains Guardrails, and no captured source mentions the people who lead it. The GitHub release notes list contributors by handle only.
OpenAI’s adjacent safety work is documented at the company level. OpenAI developed gpt-oss-safeguard with Discord, SafetyKit, and ROOST, and ROOST describes that release as one of the core components of OpenAI’s safety stack. No captured source links that work to the Guardrails maintainers, so it counts as OpenAI’s safety record rather than that team’s.
| Id | Source | Tier | Accessed |
|---|---|---|---|
| f1 | OpenAI: OpenAI Guardrails | official | 2026-09-16 |
| f2 | OpenAI closes funding round at an $852 billion valuation | press | 2026-09-16 |
| f3 | AI Defense Matrix Catalog entry | other | 2026-06-10 |
| f4 | AI Defense Matrix Catalog mapping | other | 2026-09-16 |
| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | GitHub: openai-guardrails-python repository README “Add configurable safety and compliance checks to LLM applications. OpenAI Guardrails wraps the OpenAI Python client to validate inputs and outputs, and integrates with the OpenAI Agents SDK.” | official | 2026-09-16 |
| s2 | OpenAI: Guardrails configuration wizard “Detects attempts to jailbreak LLM calls via role-playing, system prompt overrides and injections.” | official | 2026-09-16 |
| s3 | OpenAI: Guardrails Python documentation “Guardrails is a safety framework for LLM applications that automatically validates inputs and outputs using configurable checks.” | official | 2026-09-16 |
| s4 | OpenAI: Guardrails TypeScript documentation “Guardrails is a safety framework for LLM applications that automatically validates inputs and outputs using configurable checks.” | official | 2026-09-16 |
| s5 | PyPI: openai-guardrails package page “This is the Python version of OpenAI Guardrails, a package for adding configurable safety and compliance guardrails to LLM applications.” | official | 2026-09-16 |
| s6 | GitHub: openai-guardrails-python releases page “0.3.3 (2026-09-10) Bug Fixes bound secret-key extension matching work” | official | 2026-09-16 |
| s7 | OpenAI: Guardrails documentation, Prompt Injection Detection check “Detects prompt injection attempts in function calls and function call outputs using LLM-based analysis.” | official | 2026-09-16 |
| s8 | OpenAI: Guardrails documentation, Jailbreak Detection check “Analyzes text for jailbreak attempts using LLM-based detection, identifies various attack patterns, and provides confidence scores for detected attempts.” | official | 2026-09-16 |
| s9 | OpenAI: Guardrails documentation, Evaluation Tool “Evaluate guardrail performance against labeled datasets with precision, recall, F1 metrics and benchmarking capabilities.” | official | 2026-09-16 |
| s10 | OpenAI: API deprecations page “On June 3, 2026, we notified developers using Agent Builder that the product is being deprecated. ChatKit remains available.” | official | 2026-09-16 |
| s11 | OpenAI: Safety in building agents guide “Sanitize incoming inputs using built-in guardrails to redact personally identifiable information (PII) and detect jailbreak attempts.” | official | 2026-09-16 |
| s12 | GitHub: gpt-oss-safeguard repository README “gpt-oss-safeguard-120b and gpt-oss-safeguard-20b are safety reasoning models built-upon gpt-oss .” | official | 2026-09-16 |
| s13 | HiddenLayer: research post on bypassing OpenAI Guardrails “On October 6th, OpenAI released its Guardrails safety framework , a collection of heavily customizable validation pipelines that can be used to detect, filter, or block potentially harmful model inputs, outputs, and tool calls.” | research | 2026-09-16 |
| s14 | Zenity Labs: research post on AgentKit's guardrails “A deep dive into OpenAI's AgentKit guardrails, how they are implemented, and where they fail” | research | 2026-09-16 |
| s15 | Help Net Security: article on gpt-oss-safeguard “Both are fine-tuned versions of the gpt-oss open models and available under the Apache 2.0 license, allowing anyone to use, modify, and deploy them freely.” | press | 2026-09-16 |
| s16 | ROOST: blog post on the gpt-oss-safeguard release “After months of joint technical work between ROOST and OpenAI, together we are releasing one of the core components of OpenAI’s safety stack in the open—granting everyone the access to study a moderation model currently used in some of the most consequential products for safety.” | other | 2026-09-16 |
| s17 | CISA: alert on joint guidance for deploying AI systems securely “Provide methodologies and controls to protect, detect, and respond to malicious activity against AI systems and related data and services.” | regulatory | 2026-09-16 |
| s18 | NIST: pre-deployment evaluation of OpenAI's o1 model “The U.S. Artificial Intelligence Safety Institute (US AISI) and the UK Artificial Intelligence Safety Institute (UK AISI) conducted a joint pre-deployment evaluation of OpenAI’s latest model, o1 (released December 5, 2024).” | research | 2026-09-16 |
| Id | Source | Tier | Accessed |
|---|---|---|---|
| s1 | GitHub: openai-guardrails-python repository README “Add configurable safety and compliance checks to LLM applications. OpenAI Guardrails wraps the OpenAI Python client to validate inputs and outputs, and integrates with the OpenAI Agents SDK.” | official | 2026-09-16 |
| s2 | OpenAI: Guardrails configuration wizard “Detects attempts to jailbreak LLM calls via role-playing, system prompt overrides and injections.” | official | 2026-09-16 |
| s3 | OpenAI: Guardrails Python documentation “Guardrails is a safety framework for LLM applications that automatically validates inputs and outputs using configurable checks.” | official | 2026-09-16 |
| s4 | OpenAI: Guardrails TypeScript documentation “Guardrails is a safety framework for LLM applications that automatically validates inputs and outputs using configurable checks.” | official | 2026-09-16 |
| s5 | PyPI: openai-guardrails package page “This is the Python version of OpenAI Guardrails, a package for adding configurable safety and compliance guardrails to LLM applications.” | official | 2026-09-16 |
| s6 | GitHub: openai-guardrails-python releases page “0.3.3 (2026-09-10) Bug Fixes bound secret-key extension matching work” | official | 2026-09-16 |
| s7 | OpenAI: Guardrails documentation, Prompt Injection Detection check “Detects prompt injection attempts in function calls and function call outputs using LLM-based analysis.” | official | 2026-09-16 |
| s8 | OpenAI: Guardrails documentation, Jailbreak Detection check “Analyzes text for jailbreak attempts using LLM-based detection, identifies various attack patterns, and provides confidence scores for detected attempts.” | official | 2026-09-16 |
| s9 | OpenAI: Guardrails documentation, Evaluation Tool “Evaluate guardrail performance against labeled datasets with precision, recall, F1 metrics and benchmarking capabilities.” | official | 2026-09-16 |
| s10 | OpenAI: API deprecations page “On June 3, 2026, we notified developers using Agent Builder that the product is being deprecated. ChatKit remains available.” | official | 2026-09-16 |
| s11 | OpenAI: Safety in building agents guide “Sanitize incoming inputs using built-in guardrails to redact personally identifiable information (PII) and detect jailbreak attempts.” | official | 2026-09-16 |
| s12 | GitHub: gpt-oss-safeguard repository README “gpt-oss-safeguard-120b and gpt-oss-safeguard-20b are safety reasoning models built-upon gpt-oss .” | official | 2026-09-16 |
| s13 | HiddenLayer: research post on bypassing OpenAI Guardrails “On October 6th, OpenAI released its Guardrails safety framework , a collection of heavily customizable validation pipelines that can be used to detect, filter, or block potentially harmful model inputs, outputs, and tool calls.” | research | 2026-09-16 |
| s14 | Zenity Labs: research post on AgentKit's guardrails “A deep dive into OpenAI's AgentKit guardrails, how they are implemented, and where they fail” | research | 2026-09-16 |
| s15 | Help Net Security: article on gpt-oss-safeguard “Both are fine-tuned versions of the gpt-oss open models and available under the Apache 2.0 license, allowing anyone to use, modify, and deploy them freely.” | press | 2026-09-16 |
| s16 | ROOST: blog post on the gpt-oss-safeguard release “After months of joint technical work between ROOST and OpenAI, together we are releasing one of the core components of OpenAI’s safety stack in the open—granting everyone the access to study a moderation model currently used in some of the most consequential products for safety.” | other | 2026-09-16 |
| s17 | CISA: alert on joint guidance for deploying AI systems securely “Provide methodologies and controls to protect, detect, and respond to malicious activity against AI systems and related data and services.” | regulatory | 2026-09-16 |
| s18 | NIST: pre-deployment evaluation of OpenAI's o1 model “The U.S. Artificial Intelligence Safety Institute (US AISI) and the UK Artificial Intelligence Safety Institute (UK AISI) conducted a joint pre-deployment evaluation of OpenAI’s latest model, o1 (released December 5, 2024).” | research | 2026-09-16 |
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.