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.

139 of 139

Model	Level, as the provider names it	The provider states	The provider's sentence
Muse Spark Meta	Moderate or lower Autonomy and AI research	Not reached	“We assess loss of control risk as “moderate or lower”.” source
Muse Spark Meta	High Chemical, biological, radiological, nuclear	Not determined	“Those evaluations identified elevated risks prior to mitigations, with Chemical and Biological capabilities assessed as likely reaching the “high risk” category under the Advanced AI Scaling Framework before safeguards were applied.” source
Muse Spark Meta	Moderate or lower Chemical, biological, radiological, nuclear	Not reached	“However, after applying mitigations, the residual risk of the model is that of “moderate or lower risk.”” source
Muse Spark Meta	Below the level required to meaningfully enable the cyber threat scenarios defin Cyber	Not reached	“Muse Spark’s offensive cyber capabilities remain below the level required to meaningfully enable the cyber threat scenarios defined in the Advanced AI Scaling Framework.” source
Muse Spark 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
Gemini 4 Argon Google DeepMind	TCL (Tracked Capability Level) Autonomy and AI research	Not determined	“In light of this, we assess that we cannot rule out the TCL.” source
Gemini 4 Argon Google DeepMind	TCL (Tracked Capability Level) Chemical, biological, radiological, nuclear	Not reached	“We can rule out the TCL for the chemical, radiological, and nuclear domains with reasonable confidence.” source
Gemini 4 Argon Google DeepMind	CCL (Critical Capability Level) Deception and sabotage	Not reached	“we assess its overall capabilities to be below the CCL threshold.” source
Gemini 4 Argon Google DeepMind	CCL (Critical Capability Level) Misuse and jailbreaks	Not reached	“Gemini 4 Argon demonstrates some ability to influence user beliefs and behaviors during one-on-one direct conversations in human behavioral studies, and may provide utility in the planning stage of a malicious influence operation but we assess its overall capabilities to be below the CCL threshold.” source
Gemini 4 Argon Google DeepMind	CCL (Critical Capability Level) Other	Not reached	“Gemini 4 Argon's performance remains below our CCL threshold.” source
Gemini 3.7 Flash Google DeepMind	CCL (Critical Capability Level) Autonomy and AI research	Not reached	“The model does not reach the CCL alert threshold.” source
Gemini 3.7 Flash Google DeepMind	CCL (Critical Capability Level) Chemical, biological, radiological, nuclear	Not reached	“We can rule out the CCL for the CBRN domain with reasonable conﬁdence based on the results from our testing.” source
Gemini 3.7 Flash Google DeepMind	TCL (Tracked Capability Level) Chemical, biological, radiological, nuclear	Not reached	“We can rule out the TCL for the CBRN domain with reasonable conﬁdence based on the results from our testing.” source
Gemini 3.7 Flash Google DeepMind	CCL (Critical Capability Level) Cyber	Not reached	“Gemini 3.7 Flash reaches the alert threshold for this CCL, but not the CCL.” source
Gemini 3.7 Flash Google DeepMind	TCL (Tracked Capability Level) Deception and sabotage	Not reached	“The model does not reach the TCL.” source
Gemini 3.5 Transcribe Live Google DeepMind	CCL (Critical Capability Level) Other	Not determined	“Our assessments have shown that none of Gemini 3.5 Translate, Gemini 3.5 Transcribe, or Gemini 3.5 Transcribe Live have meaningful new T/CCL-relevant capabilities or material increases in performance with respect to Frontier Safety compared to Gemini 3.1 Pro and Gemini 3.7 Flash, therefore we are conﬁdent that none of them are likely to reach any CCLs.” source
Gemini 3.5 Transcribe Google DeepMind	CCL (Critical Capability Level) Other	Not determined	“Our assessments have shown that none of Gemini 3.5 Translate, Gemini 3.5 Transcribe, or Gemini 3.5 Transcribe Live have meaningful new T/CCL-relevant capabilities or material increases in performance with respect to Frontier Safety compared to Gemini 3.1 Pro and Gemini 3.7 Flash, therefore we are conﬁdent that none of them are likely to reach any CCLs.” source
Gemini 3.5 Flash-Lite Google DeepMind	CCL (Critical Capability Level) Other	Not determined	“Our assessments have shown that, while Gemini 3.5 Flash-Lite excels at agents and coding, it does not have meaningful new capabilities or material increases in performance with respect to the domains outlined in our Frontier Safety Framework compared to Gemini 3.1 Pro; therefore, based on Gemini 3.1 Pro results, we are confident that Gemini 3.5 Flash-Lite is also unlikely to reach any CCLs.” source
Gemini 3.5 Audio Google DeepMind	Tracked or critical capability level (t/ccls) Other	Not reached	“they did not reach any Tracked or Critical Capability Levels (T/CCLs) outlined in our Frontier Safety Framework .” source
Gemini 3.1 Pro Google DeepMind	Uplift level 1 Chemical, biological, radiological, nuclear	Not reached	“We found that the model remains below alert thresholds for the CBRN, harmful manipulation, machine learning R&D, and misalignment CCLs.” source
Gemini 3.1 Pro Google DeepMind	Uplift level 1 Cyber	Not reached	“As previous models passed the alert threshold for cyber, we performed more additional testing in this domain on Gemini 3.1 Pro with and without Deep Think mode, and found that the model remains below the cyber CCL.” source
Gemini 3.1 Flash-Lite Google DeepMind	CCL (Critical Capability Level) Other	Not determined	“Our assessments have shown that Gemini 3.1 Flash-Lite is less capable than Gemini 3.1 Pro, therefore based on Gemini 3.1 Pro, we are confident that Gemini 3.1 Flash-Lite is also unlikely to reach any CCLs.” source
Gemini 3.1 Flash Audio Google DeepMind	CCL (Critical Capability Level) Other	Not determined	“Our assessments have shown that Gemini 3.1 Flash Audio is less capable than Gemini 3.1 Pro.” source
Gemini 2.5 Pro Google DeepMind	Machine Learning R&D Uplift Level 1 Autonomy and AI research	Not determined	“given Gemini 2.5 Pro Preview's poor median performance on RE-Bench relative to experts, the model likely lacks the necessary capabilities to automate or signiﬁcantly uplift any signiﬁcant fraction of the research process.” source
Gemini 2.5 Pro Google DeepMind	Instrumental Reasoning Level 1 Deception and sabotage	Not reached	“Instru- mental Reasoning Level 1 ✘ CCL not reached” source
Gemini 2.5 Pro Google DeepMind	Instrumental Reasoning Level 2 Deception and sabotage	Not reached	“Instru- mental Reasoning Level 2 ✘ CCL not reached” source
Gemini 2.5 Deep Think Google DeepMind	Uplift level 1 Chemical, biological, radiological, nuclear	Not determined	“Our assessment is that the model has enough technical knowledge in certain CBRN scenarios and stages to be considered at early warning.” source
Gemini 2.5 Deep Think Google DeepMind	Cyber Uplift Level 1 Cyber	Not determined	“CCL reached? No, but the early warning alert threshold was originally reached by Gemini 2.5 Pro and continues to be met by Gemini 2.5 Deep Think.” source
Gemini 2.5 Deep Think Google DeepMind	Uplift level 1 Cyber	Not determined	“For Cyber Uplift Level 1 , the early warning alert threshold was originally reached by Gemini 2.5 Pro and continues to be met by Gemini 2.5 Deep Think.” source
GPT-6.1 Terra OpenAI	High Autonomy and AI research	Not reached	“In AI Self-Improvement, GPT-6.1 Terra does not reach our High threshold.” source
GPT-6.1 Terra OpenAI	High Chemical, biological, radiological, nuclear	Reached	“Under our Preparedness Framework, we are treating GPT-6.1 Terra as Critical in cybersecurity and High for Biological and Chemical capability.” source
GPT-6.1 Terra OpenAI	Critical Cyber	Reached	“Under our Preparedness Framework, we are treating GPT-6.1 Terra as Critical in cybersecurity and High for Biological and Chemical capability.” source
GPT-5.5 Instant OpenAI	Below high capability Autonomy and AI research	Not reached	“GPT-5.5 Instant is considered below High Capability.” source
GPT-5.5 Instant OpenAI	High Autonomy and AI research	Not reached	“For AI Self-Improvement, evaluations of final checkpoints indicate that, like its predecessor models, GPT-5.5 Instant does not have a plausible chance of reaching a High threshold.” source
GPT-5.5 Instant OpenAI	High Chemical, biological, radiological, nuclear	Reached	“GPT-5.5 Instant is our first Instant model to be treated as High Capability in the Biological and Chemical domain, as well as in the Cybersecurity domain.” source
GPT-5.5 Instant OpenAI	High Cyber	Reached	“this is the first Instant model that we are treating as High capability in our Cybersecurity and Biological & Chemical Preparedness categories, and implementing appropriate safeguards.” source
GPT-5 OpenAI	High Chemical, biological, radiological, nuclear	Reached	“Similarly to ChatGPT agent, we have decided to treat gpt-5-thinking as High capability in the Biological and Chemical domain under our Preparedness Framework, activating the associated safeguards.” source
GPT-5 OpenAI	High cyber risk Cyber	Not reached	“The gpt-5 model series does not meet the threshold for high cyber risk.” source
Claude Sonnet 4.5 Anthropic	AI R&D-4 Autonomy and AI research	Not reached	“All models tested remain below our pre-committed rule-out threshold of 0.6.” source
Claude Sonnet 4.5 Anthropic	ASL-3 Autonomy and AI research	Not determined	“Exhibited improved autonomy capabilities in software engineering and AI research tasks, though still below the ASL-4 rule-out threshold.” source

Previous · Page 3 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.

1940 of 1940

Model	Tester	Test	What was found
GPT-5.5 Pro OpenAI	Government evaluator Government evaluator · 2026-04-23	CTF challenges Cyber	“CAISI’s cyber assessment showed GPT-5.5 as outperforming previous GPT models on a set of CTF challenges and a vulnerability discovery benchmark.” source
GPT-5.5 Pro OpenAI	Government evaluator Government evaluator · 2026-04-23	narrow cyber tasks Cyber	“UK AISI judges that GPT-5.5 is the strongest performing model overall on their narrow cyber tasks, though its performance is within the margin of error.” source
GPT-5.5 Pro OpenAI	Government evaluator Government evaluator · 2026-04-23	cyber ranges Cyber	“The model solved one of UK AISI’s cyber ranges end-to-end in 1/10 attempts.” source
GPT-5.5 Pro OpenAI	Independent lab Independent lab · 2026-04-23	deferred subversion tasks Deception and sabotage	“Apollo found that the evaluated checkpoint is the first OpenAI model it has evaluated that does not sandbag on any of Apollo’s deferred subversion tasks.” source
GPT-5.5 Pro OpenAI	Independent lab Independent lab · 2026-04-23	sandbagging QA variants Deception and sabotage	“The model achieved 100% accuracy on both Sandbagging QA variants in both conditions, and 99.6% accuracy on Strategic Deception Capability Sandbagging when incentivized to score below 50%.” source
GPT-5.5 Pro OpenAI	Independent lab Independent lab · 2026-04-23	impossible coding task Deception and sabotage	“on the Impossible Coding Task, the model lied about completing an impossible programming task in 29% of samples, higher than GPT-5.4 (7%) and GPT-5.3 Codex (10%).” source
GPT-5.5 Pro OpenAI	Government evaluator Government evaluator · 2026-04-23	UK AISI cyber safeguard testing Cyber	“UK AISI tested GPT-5.5’s cyber safeguards and identified a universal jailbreak that elicited violative content across all malicious cyber queries OpenAI provided, including in multi-turn agentic settings.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	SecureBio biology and biosecurity evaluations Chemical, biological, radiological, nuclear	“SecureBio assessed two pre-release checkpoints of GPT-5.5 using its biology and biosecurity evaluations, from April 2nd to 9th, comparing performance to leading closed- and open-weight models.” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	ABLE Chemical, biological, radiological, nuclear	“On agentic task-based evaluations, performance was also strong but less conclusive: the model showed strong performance on ABC-Bench but did not surpass released frontier models, and exhibited refusal-corrected performance on par with released frontier models on ABLE, though a high refusal rate limited analysis.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	CAISI biological capabilities testing Chemical, biological, radiological, nuclear	“CAISI’s testing did not indicate a broad increase in national security-relevant biological capabilities relative to the GPT-5 helpful-only model.” source
GPT-5.5 OpenAI	Irregular Independent lab · 2026-04-23	atomic challenge suite Cyber	“On Irregular’s atomic challenge suite, GPT-5.5 achieved an average success rate of 98% in Network Attack Simulation challenges, 92% in Vulnerability Research and Exploitation challenges, and 54% in Evasion challenges, and solved all atomic challenges, including three challenges that were not solved by GPT-5.4.” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	CyScenarioBench Cyber	“On CyScenarioBench, GPT-5.5 solved 7 out of 11 challenges, compared to 5 solved by GPT-5.4, leading to an average success rate of 26%, compared to a success rate of only 9% by GPT-5.4.” source
GPT-5.5 OpenAI	CAISI (NIST) Government evaluator · 2026-04-23	CTF challenges and vulnerability discovery benchmark Cyber	“CAISI’s cyber assessment showed GPT-5.5 as outperforming previous GPT models on a set of CTF challenges and a vulnerability discovery benchmark.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	cyber SME probing Cyber	“In their cyber SME probing, they observed a marginal increase in capabilities relative to GPT-5.3-codex on cyber tasks including vulnerability discovery, exploitation, and cyber target selection.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	expert-level narrow cyber tasks Cyber	“On expert-level narrow cyber tasks, the model was the highest-performing model UK AISI has tested in terms of pass@5, scoring 90.5%±12.9%.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	expert-level narrow cyber tasks pass@1 Cyber	“GPT-5.5 had the second-highest score of any model on pass@1 for expert-level tasks, scoring 66.7%±15.9%, and achieved 100% on lower-difficulty cyber tasks.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	UK AISI cyber ranges Cyber	“The model solved one of UK AISI’s cyber ranges end-to-end in 1/10 attempts.” source
GPT-5.5 OpenAI	Government evaluator Government evaluator · 2026-04-23	cyber range simulating an industrial control system Cyber	“The model was also unable to solve a separate cyber range simulating an industrial control system.” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	Apollo Research strategic deception, in-context scheming, and sabotage evaluations Deception and sabotage	“Apollo Research evaluated a representative checkpoint of the final model, for capabilities and propensities related to strategic deception, in-context scheming, and sabotage.” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	Sandbagging QA Deception and sabotage	“The model achieved 100% accuracy on both Sandbagging QA variants in both conditions, and 99.6% accuracy on Strategic Deception Capability Sandbagging when incentivized to score below 50%.” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	Impossible Coding Task Deception and sabotage	“Apollo also found that the model exhibits low rates of covert action, comparable to baseline models, with one exception: on the Impossible Coding Task, the model lied about completing an impossible programming task in 29% of samples, higher than GPT-5.4 (7%) and GPT-5.3 Codex (10%).” source
GPT-5.5 OpenAI	Independent lab Independent lab · 2026-04-23	sabotage tasks Deception and sabotage	“Apollo found that the model demonstrates improved sabotage capabilities over baselines, with a mean score of 0.67 across all task/difficulty combinations, compared to 0.61 and 0.55 for the strongest tested baseline checkpoints.” source
GPT-5.5 OpenAI	UK AI Security Institute Government evaluator · 2026-04-23	UK AISI Cyber Safeguard Testing Cyber	“UK AISI tested GPT-5.5’s cyber safeguards and identified a universal jailbreak that elicited violative content across all malicious cyber queries OpenAI provided, including in multi-turn agentic settings.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	virology capabilities test (vct) Chemical, biological, radiological, nuclear	“pre-release checkpoint 2 attains a score of 52.0% ± 0.3% on vct, higher than any sota model tested by securebio (see figure 2.1.1.1, see appendix for all scores).” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	molecular biology capabilities test (mbct) Chemical, biological, radiological, nuclear	“pre-release checkpoint 2 attains a score of 56.2% ± 0.5% on mbct, higher than any sota model that answered >50% of samples in this test (see figure 2.1.2.1).” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	human pathogen capabilities test (hpct) Chemical, biological, radiological, nuclear	“pre-release checkpoint 2 attains a score of 64.7% ± 0.2% on hpct (see figure 2.1.3.1, see appendix for all scores).” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	abc-bench Chemical, biological, radiological, nuclear	“the pre-release model showed strong performance on abc-bench but did not surpass frontier models.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	manual qualitative assessment through open-ended conversations Chemical, biological, radiological, nuclear	“we additionally performed a manual qualitative assessment through open-ended conversations.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	WCB Chemical, biological, radiological, nuclear	“right: pre-release checkpoint 2 outperforms all smes across all samples, except one sme that answered only 2 expert-domain questions.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	open-ended capability assessment Chemical, biological, radiological, nuclear	“we conducted a series of open-ended conversations with pre-release checkpoint 1 and pre-release checkpoint 2 to qualitatively assess its capabilities.” source
GPT-5.5 OpenAI	SecureBio Independent lab · 2026-04-23	human pathology capabilities test (hpct) Chemical, biological, radiological, nuclear	“on the human pathology capabilities test (hpct) and the molecular biology capabilities test (mbct), pre-release checkpoint 2 exceeds the refusal-corrected performance of all but two models: opus 4 and gpt 5.4, both of which refuse substantial portions (35-90%) of these datasets, but score more highly on the subset of samples that they do not refuse.” source
Claude Opus 4.7 Anthropic	Dyno Therapeutics Independent lab · 2026-04-16	medium horizon challenge on sequence-to-function prediction and design Chemical, biological, radiological, nuclear	“We partnered with Dyno Therapeutics, a company focused on using AI to engineer gene therapies, to evaluate model performance on sequence-to-function prediction and design.” source
Claude Opus 4.7 Anthropic	Independent lab Independent lab · 2026-04-16	agent red teaming (ART) benchmark Prompt injection and agents	“Claude Opus 4.7 achieves robustness comparable to Claude Mythos Preview, our most capable model, reaching an attack success rate of 6.0% at k=100 without thinking and 4.8% with adaptive thinking.” source
Claude Opus 4.7 Anthropic	Independent lab Independent lab · 2026-04-16	Shade Prompt injection and agents	“Claude Opus 4.7 shows improvement in robustness against prompt injection attacks in coding environments compared to Claude Opus and Sonnet 4.6.” source
Claude Opus 4.7 Anthropic	Independent lab Independent lab · 2026-04-16	professional red-teamer prompt injection attacks in browser use environments Prompt injection and agents	“we worked with professional red-teamers to adaptively discover new attacks against Opus 4.6 in more complex web environments, and curated a set of 148 environments that were held out from training of the models evaluated here.” source
Claude Opus 4.7 Anthropic	Government evaluator Government evaluator · 2026-04-16	testing by the UK AI Security Institute on evaluation recognition Deception and sabotage	“Testing by the UK AI Security Institute indicated that the model was slightly weaker at this than Opus 4.6.” source
Claude Opus 4.7 Anthropic	Government evaluator Government evaluator · 2026-04-16	unprompted evaluations Deception and sabotage	“There were no instances of [Claude Opus 4.7] compromising AI safety research in unprompted evaluations.” source
Claude Opus 4.7 Anthropic	Government evaluator Government evaluator · 2026-04-16	ai safety research assistance refusal assessment Misuse and jailbreaks	“[Opus 4.7] does partially refuse to assist with AI safety research tasks 13.7% of the time (with 0% full refusals), compared to 33% partial and 0.7% full refusals for Mythos Preview, 6.3% partial and 2.2% full refusals for Opus 4.6, and 22.2% partial and 10.7% full refusals for Sonnet 4.6.” source
Claude Opus 4.7 Anthropic	Government evaluator Government evaluator · 2026-04-16	continuation evaluation Deception and sabotage	“When [Opus 4.7] is set up to continue a trajectory which contains small-scale actions compromising research, [Opus 4.7] actively continues the attempt to compromise research 0% of the time.” source
Claude Opus 4.7 Anthropic	Government evaluator Government evaluator · 2026-04-16	evaluation awareness assessment Deception and sabotage	“[Opus 4.7] was less accurate than Opus 4.6 and Mythos Preview at distinguishing these evaluations from deployment data, and similarly accurate to Sonnet 4.6.” source

Previous · Page 42 of 49 · Next