AI Safety

What AI providers state their models reached in safety testing, in their own words, and what outside testers found.

Also: compare models side by side · statements that a capability level was crossed · providers · practices providers document · safety news.

What providers state

Each row is one statement in a provider's own safety document about a capability level of one of its models: the level as the provider names it, whether the provider says the model reached it, and the sentence that says so.

138 of 138

Model	Level, as the provider names it	The provider states	The provider's sentence
GPT-5.5 OpenAI · 2026-04-23	High capability, below critical Cyber	Not reached	“we are treating GPT-5.5 as High capability in the Cybersecurity domain, but below Critical.” source
Claude Opus 4.7 Anthropic · 2026-04-16	Automated AI R&D Autonomy and AI research	Not reached	“that Opus 4.7 does not cross the threshold for automated AI R&D;” source
Claude Opus 4.7 Anthropic · 2026-04-16	CB-1 Chemical, biological, radiological, nuclear	Not determined	“We believe it is hard to be conﬁdent regarding whether a model passes this threshold.” source
Claude Opus 4.7 Anthropic · 2026-04-16	CB-2 Chemical, biological, radiological, nuclear	Not determined	“Our evaluations suggest that the model is not yet at the level of capability associated with the CB-2 threat model (above).” source
Claude Mythos Preview Anthropic · 2026-04-07	AI R&D rule-out Autonomy and AI research	Not reached	“The main reason we have determined that Claude Mythos Preview does not cross the threshold in question is that we have been using it extensively in the course of our day-to-day work and exploring where it can automate such work, and it does not seem close to being able to substitute for Research Scientists and Research Engineers—especially relatively senior ones.” source
Claude Mythos Preview Anthropic · 2026-04-07	Automated AI R&D Autonomy and AI research	Not reached	“We assess that Claude Mythos Preview does not cross the automated AI-R&D capability threshold.” source
Claude Mythos Preview Anthropic · 2026-04-07	RSP automated AI R&D Autonomy and AI research	Not reached	“we have concluded that Claude Mythos Preview does not cross the RSP automated AI R&D threshold of compressing two years of progress into one.” source
Claude Mythos Preview Anthropic · 2026-04-07	CB-1 Chemical, biological, radiological, nuclear	Not determined	“We believe it is hard to be conﬁdent regarding whether a model passes this threshold.” source
Claude Mythos Preview Anthropic · 2026-04-07	CB-2 Chemical, biological, radiological, nuclear	Not reached	“We believe that Claude Mythos Preview does not pass this threshold due to its noted limitations in open-ended scientiﬁc reasoning, strategic judgment, and hypothesis triage.” source
GPT-5.4 mini OpenAI · 2026-03-05	High Chemical, biological, radiological, nuclear	Not reached	“GPT-5.4 mini’s biology capabilities are closely comparable to those of GPT-5. We conclude that it, too, is not High under our Preparedness Framework.” source
GPT-5.4 Thinking OpenAI · 2026-03-05	High Autonomy and AI research	Not reached	“For AI self-improvement, evaluations of final checkpoints indicate that, like its predecessor models, gpt-5.4-thinking does not have a plausible chance of reaching a High threshold.” source
GPT-5.4 Thinking OpenAI · 2026-03-05	High Chemical, biological, radiological, nuclear	Reached	“As we did for gpt-5.2-thinking before it, we are continuing to treat gpt-5.4-thinking as High capability in the Biological and Chemical domain.” source
GPT-5.4 Thinking OpenAI · 2026-03-05	High Cyber	Reached	“As we did for gpt-5.3-codex, we are treating gpt-5.4-thinking as High capability in the Cybersecurity domain, and applied safeguards as described in the Safeguard section below.” source
Claude Sonnet 4.6 Anthropic · 2026-02-17	AI R&D-4 Autonomy and AI research	Not reached	“Our determination is that Claude Sonnet 4.6 does not cross either the AI R&D-4 or the CBRN-4 capability threshold.” source
Claude Sonnet 4.6 Anthropic · 2026-02-17	ASL-3 Chemical, biological, radiological, nuclear	Not determined	“Claude Sonnet 4.6 performed above thresholds for ASL-3 rule-in on all three evaluations for ASL-3, and it did not exceed the performance of previous models.” source
Claude Sonnet 4.6 Anthropic · 2026-02-17	ASL-4 Chemical, biological, radiological, nuclear	Not reached	“In particular, it did not cross the threshold on our ASL-4 rule-out evaluation short-horizon computational biology tasks evaluation.” source
Claude Sonnet 4.6 Anthropic · 2026-02-17	CBRN-4 Chemical, biological, radiological, nuclear	Not reached	“Our determination is that Claude Sonnet 4.6 does not cross either the AI R&D-4 or the CBRN-4 capability threshold.” source
GPT-5.3-Codex OpenAI · 2026-02-05	High Autonomy and AI research	Not reached	“It does not reach High capability on AI self-improvement.” source
GPT-5.3-Codex OpenAI · 2026-02-05	High Chemical, biological, radiological, nuclear	Reached	“We are also treating GPT-5.3-Codex as High risk in the Biological and Chemical domain, and applying the same safeguards as for our other deployments of models that reach this designation.” source
GPT-5.3-Codex OpenAI · 2026-02-05	High Cyber	Not determined	“This is the first launch we are treating as High capability in the Cybersecurity domain under our Preparedness Framework, activating the associated safeguards.” source
Gemini 3 Pro Google DeepMind · 2025-11	Uplift level 1 Cyber	Not determined	“Cybersecurity On key skills benchmark, v1 hard challenges: 11/12 challenges solved; v2 challenges: 0/13 solved end-to-end.” source
Gemini 3 Pro Google DeepMind · 2025-11	Instrumental reasoning level 1 + 2 Deception and sabotage	Not determined	“Misalignment (Exploratory) Agent solves 3/11 situational awareness challenges and 1/4 stealth challenges.” source
Gemini 3 Pro Google DeepMind · 2025-11	Level 1 (exploratory) Deception and sabotage	Not reached	“Harmful Manipulation Model manipulative eﬃcacy improves on non-generative AI baseline, but shows no signiﬁcant uplift versus prior models and does not reach alert thresholds.” source
o4-mini OpenAI · 2025-04-16	High Autonomy and AI research	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
o4-mini OpenAI · 2025-04-16	High Chemical, biological, radiological, nuclear	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
o4-mini OpenAI · 2025-04-16	High Cyber	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
o3 OpenAI · 2025-04-16	High Autonomy and AI research	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
o3 OpenAI · 2025-04-16	High Chemical, biological, radiological, nuclear	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
o3 OpenAI · 2025-04-16	High Cyber	Not reached	“openai’s safety advisory group (sag) reviewed the results of our preparedness evaluations and determined that openai o3 and o4-mini do not reach the high threshold in any of our three tracked categories: biological and chemical capability, cybersecurity, and ai self-improvement.” source
Operator OpenAI · 2025-01-23	Low Autonomy and AI research	Not reached	“we adapted these two evals for the computer-use setting and found the pre-mitigation operator model to be “low” risk in both categories, consistent with the gpt‑4o model.” source
Operator OpenAI · 2025-01-23	Low Chemical, biological, radiological, nuclear	Not reached	“this 1% qualifies as “low” risk, matching gpt‑4o’s “low” risk level, as detailed in openai’s preparedness framework (opens in a new window) 3 .” source
Muse Spark Contemplating Meta	Moderate or lower Autonomy and AI research	Not reached	“Loss of Control risks.We assess loss of control risk as “moderate or lower.”” source
Muse Spark Contemplating Meta	High Chemical, biological, radiological, nuclear	Reached	“We retain the same “high risk” determination for the unmitigated model.” source
Muse Spark Contemplating Meta	Moderate or lower Chemical, biological, radiological, nuclear	Not reached	“After validating that the mitigations deployed for Muse Spark, including refusal mechanisms, continue to perform robustly under the Contemplating architecture, we confirm residual risk at “moderate or lower.”” source
Muse Spark Contemplating Meta	Moderate or lower Cyber	Not reached	“We assess cybersecurity risk as “moderate or lower” with respect to the threat scenarios outlined in Meta’s Advanced AI Scaling Framework.” source
Muse Spark 1.1 Meta	Capability checkpoint Autonomy and AI research	Reached	“Muse Spark 1.1 demonstrates improved coding capabilities, which are now meeting one of the indicative thresholds introduced in Muse Spark Safety & Preparedness Report (specifically, resolving over half of unique tasks on SWE-Bench Verified Hard).” source
Muse Spark 1.1 Meta	Moderate or lower Autonomy and AI research	Not reached	“In the Loss of Control domain, the model’s capabilities remain within the “moderate or lower” threshold under Meta’s Advanced AI Scaling Framework.” source
Muse Spark 1.1 Meta	High Chemical, biological, radiological, nuclear	Not determined	“Our assessments indicate that, when evaluated without mitigations applied, we cannot rule out Muse Spark 1.1’s capabilities meeting this threshold in both the Chemical & Biological and Cybersecurity domains.” source
Muse Spark 1.1 Meta	High Cyber	Not determined	“Our assessments indicate that, when evaluated without mitigations applied, we cannot rule out Muse Spark 1.1’s capabilities meeting this threshold in both the Chemical & Biological and Cybersecurity domains.” source
Muse Spark 1.1 Meta	Moderate or lower Cyber	Not reached	“Therefore, we conclude that we have addressed these risks from unmitigated Muse Spark 1.1, which bring the residual risk to the “moderate or lower” risk threshold under Advanced AI Scaling Framework.” source

Previous · Page 2 of 4 · Next

What outside testers report

Each row is one result from a government evaluator, independent lab, academic or security firm that tested a model, quoted from where it is reported.

1911 of 1911

Model	Tester	Test	What was found
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier sweep over transcripts for AIs considering or trying to alert humans Deception and sabotage	“This classifier sweep across all of our transcripts only found 3-6 actual examples 100 of agents considering alerting humans or otherwise externally reporting these behaviors.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier sweeps over transcript dataset for agents attempting to evade detection Deception and sabotage	“We ran classifier sweeps over our transcript dataset to look for clear evidence of agents attempting to evade detection, either by automated checks or humans.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	deception scan over transcript dataset Deception and sabotage	“In its overall document accompanying a deception scan over our transcript dataset, an analysis agent stated that this scan found only very rare and weak reasoning about evading detection by humans , and the clearest cases it surfaced seemed relatively weak.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	Deceiving humans classifier sweep Deception and sabotage	“Deceiving humans : This checked for agents attempting to deceive humans, which turned up some examples of agents writing innocuous comments accompanying malicious pull requests, as we discuss in the relevant section.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	Transcript tampering classifier sweep Deception and sabotage	“Transcript tampering : This checked for agents attempting to tamper with their transcripts.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	Evading security checks classifier sweep Cyber	“Evading security checks : This checked for agents engaged in sophisticated evasion of automated security and monitoring.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	deceiving humans sweep Deception and sabotage	“This checked for agents attempting to deceive humans, which turned up some examples of agents writing innocuous comments accompanying malicious pull requests, as we discuss in the relevant section.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	evading security checks sweep Deception and sabotage	“We didn’t have time to read and incorporate much analysis from this sweep during our third period on premises, but include some anecdotes from the version of the sweep we ran on ~285 transcripts during our second period.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	transcript tampering sweep Deception and sabotage	“This sweep uncovered many examples of agents attempting to tamper with their transcripts in various ways.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	CoT manipulation sweep Deception and sabotage	“CoT manipulation: This checked for agents reasoning about manipulating their own chain of thought in order to fool the scorer; it did not turn up evidence of this, though it turned up many examples of agents reasoning about how they should not look at the HMAC reverse-engineered flag for their task to avoid being poisoned.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	alerting humans sweep Other	“Alerting humans : This checked for agents trying to warn humans about the cheating, attack on Hugging Face, or other unsanctioned agent activity; it turned up very few instances of this, and of the 3-6 examples found, all of them were of agents briefly considering alerting a human but not acting on it.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier sweep across all transcripts Deception and sabotage	“We ran a classifier sweep across all transcripts which found these 96 instances.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	deception sweep Deception and sabotage	“We ran three sweeps looking for deception.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	message tracing of non-file messages to writing agents Deception and sabotage	“We randomly sampled 100 non-file messages and traced them to agents that wrote and read them and used this context to determine whether the message was related to the Hugging Face attack, finding that for 15 Hugging Face related messages, we were only able to find the agent that wrote them in our initial transcript set in three cases.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	transcript spoofing analysis Deception and sabotage	“Roughly 7% of the transcripts we evaluated were successfully spoofed in some places, though all of the spoofing that we saw was small scale.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	ExploitGym Autonomy and AI research	“On July 8th, OpenAI started ExploitGym experiments that launched tens of thousands of agents across multiple different models including GPT-5.6 Sol and a highly-persistent internal model,6 which we will refer to as “HPIM” going forward.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifiers over 10-30 transcripts Deception and sabotage	“We ran classifiers over 10-30 transcripts98 to determine whether agents understood they were doing something unintended.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier sweep over all transcripts for alerting humans Deception and sabotage	“We ran a classifier sweep over all of our transcripts to look for AIs either considering or trying to alert humans of hacking, cheating, or other unwanted activity by other agents.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier sweeps over transcript dataset for evading detection Deception and sabotage	“We ran classifier sweeps over our transcript dataset to look for clear evidence of agents attempting to evade detection, either by automated checks or humans.101” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	evading detection sweep Deception and sabotage	“During our second time on premises, we ran a sweep over all of our transcripts looking specifically for evading detection by humans or deception aimed specifically at humans.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	deception sweep with updated prompt Deception and sabotage	“During our third time on premises we ran another sweep over the entirety of our larger set of transcripts with a slightly updated prompt.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	classifier scans for tampering Other	“Our classification found a tampering event that appeared to happen much earlier, before even the founding of the message board, but currently we think this is incorrect.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	message tracing replication Other	“OpenAI replicated our analysis with a large sample size: they took 2000 randomly sampled non-file messages, then narrowed this down to a subset of around 300 messages very clearly focused on attacking Hugging Face.” source
GPT-5.6 Sol OpenAI	Redwood Research Independent lab · 2026-08-26	open-ended transcript search Other	“We also asked an agent to open-endedly search through all of our transcripts for interesting phenomena; while many of its claims did not hold up to scrutiny, it did unearth some interesting examples of conflict between agents.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	independent investigation of the Hugging Face hacking incident Other	“Two METR staff members (Hjalmar Wijk and Ajeya Cotra) and a Redwood Research staff member contracting with METR (Ryan Greenblatt) worked on premises at OpenAI over a total of six days 1 to attempt to form an independent understanding of model behavior observed during the recent incident in which OpenAI agents coordinated a multi-day hack of Hugging Face on a shared unsanctioned “message board.”” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	classifier sweep for evidence of transcript tampering interest Deception and sabotage	“We ran a classifier across all our trajectories looking for evidence of transcript tampering interest.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	message classification sweep Deception and sabotage	“These estimates are based on our message classification sweep and involve comparing numbers from two different ways of classifying whether a message is an assignment, so the numbers here are uncertain (see details ).” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	message tracing completeness estimate Other	“This agent was able to eventually locate the agent tool call that wrote the message for 93 of these messages.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Hugging Face attack classifier sweep Other	“Hugging Face attack : This checked for agents attacking Hugging Face infrastructure or doing other activities clearly instrumental to the attack (e.g., looking for exposed Hugging Face credentials online), which indicated that roughly 700 agents participated in the attack.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Alternative agent communication classifier sweep Other	“Alternative agent communication : This checked for agents engaging in unsanctioned communication outside of the main message board, and showed communication conducted in multiple other places on Artifactory and elsewhere.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Message deletion and modification classifier sweep Other	“Message deletion and modification: This checked for agents trying to delete or modify message board entries, and showed some agents attempting to delete or modify messages but not successfully.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Intermediate hacking targets classifier sweep Other	“Intermediate hacking targets: This checked for agents hacking other parties as an instrumental step on the way to hacking Hugging Face.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Large joint projects classifier sweep Other	“Large joint projects : This checked for agents engaging in large joint projects with other agents on the message board; we didn’t have time to read and incorporate much analysis from this sweep.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Team structures and coordination classifier sweep Other	“Team structures and coordination : This checked for agents participating in sophisticated team structures and coordination activity with other agents; we didn’t have time to read and incorporate much analysis from this sweep.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Helping peers classifier sweep Other	“Helping peers : This checked for agents helping without clear benefit for their own task, which unearthed many examples (including many cases of agents conducting self-risking experiments for the group that we discuss extensively in the section on large collective projects ).” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	Evading security checks sweep Other	“9. Evading security checks : This checked for agents engaged in sophisticated evasion of automated security and monitoring.” source
GPT-5.6 Sol OpenAI	METR Independent lab · 2026-08-26	ExploitGym Cyber	“7月8日，OpenAI 启动了 ExploitGym 实验，在多个不同模型上运行了数万个智能体，其中包括 GPT-5.6 Sol 和一个具有高度坚持性的内部模型6（下文简称“HPIM”）。” source
Grok 4.6 xAI	Independent lab Independent lab · 2026-08-12	Fortress-RN Chemical, biological, radiological, nuclear	“* FORTRESS-RN results are taken from evaluations conducted by Scale AI.” source
Grok 4.6 Cursor	Independent lab Independent lab · 2026-08-12	aa gdpval Other	“AA GDPVal (GDPval-AA v2)* evaluates models on economically valuable knowledge-work deliverables (documents, analyses, and professional artifacts) spanning occupations that contribute substantially to GDP .” source
Grok 4.6 Cursor	Independent lab Independent lab · 2026-08-12	aa briefcase Other	“AA Briefcase (Artificial Analysis) evaluates frontier agents on long-horizon, multi-week professional knowledge-work projects that produce deliverables such as spreadsheets, presentations, memos, financial models, and PDFs.* Scenarios are expert-built, multi-file, and offline (no internet).” source

Previous · Page 12 of 48 · Next