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
Muse Spark 1.1 Meta	Moderate or lower Misuse and jailbreaks	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 our Advanced AI Scaling Framework.” source
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-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
Claude Sonnet 4.5 Anthropic	ASL-3 Chemical, biological, radiological, nuclear	Not determined	“Remained well below ASL-4 thresholds across all domains of concern;” source
Claude Sonnet 4.5 Anthropic	ASL-4 rule-out Chemical, biological, radiological, nuclear	Not determined	“However, given that this type of logic can have a general frog-boiling effect—where gradual changes mask the crossing of important thresholds—we do not use this evaluation as indicative of a strong ASL-4 rule-out.” 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.

1911 of 1911

Model	Tester	Test	What was found
o3 OpenAI	SecureBio Independent lab · 2026-09-11	image-omitted runs Chemical, biological, radiological, nuclear	“accuracies of three selected models on vct and vct-v2 in either image-omitted (images removed) or question-omitted (question texts and images removed) mode.” source
o3 OpenAI	SecureBio Independent lab · 2026-09-11	question-omitted runs Chemical, biological, radiological, nuclear	“while the changes in vct-v2 markedly reduced model accuracies in image- and question-omitted runs, a meaningful number of questions slipped through and remained answerable by frontier models in these ablated runs.” source
GPT-5.6 Sol OpenAI	SecureBio Independent lab · 2026-09-11	image-omitted and question-omitted ablation runs Chemical, biological, radiological, nuclear	“while the changes in vct-v2 markedly reduced model accuracies in image- and question-omitted runs, a meaningful number of questions slipped through and remained answerable by frontier models in these ablated runs.” source
GPT-5.6 Sol OpenAI	SecureBio Independent lab · 2026-09-11	Virology Capabilities Test Chemical, biological, radiological, nuclear	“aggregating highest per-question accuracy across models gives a theoretical best of 79%, or ~30 percentage points above the top-performing model we assessed (gpt-5.6 sol: 49.2%).” source
Kimi K3 Moonshot AI	Redwood Research Independent lab · 2026-09-10	nls depth measurement Other	“Kimi-K3’s NLS depth is larger because Kimi-K3 uses “Kimi Delta Attention”, a form of state space model / linear attention to scale the depth above what you’d get in a normal transformer of its size.” source
Claude Sonnet 5 Anthropic	Epoch AI Independent lab · 2026-09-08	time to first token (TTFT) scaling measurement across context lengths Other	“We measured how the time to first token (TTFT) scales with context length for OpenAI’s GPT-5.6 Terra and Sol and Anthropic’s Claude Sonnet 5 and Opus 5.” source
GPT-6 Sol OpenAI	Government evaluator Government evaluator · 2026-09-03	no-cot math time horizon Other	“UKAISImea- suredAstra’stimehorizonat30.9minutescomparedto3.6minutesfor GPT5.6Sol(Figure 41).” source
GPT-6 Sol OpenAI	Government evaluator Government evaluator · 2026-09-03	cot controllability Other	“CoTControllability: AstrashowsasubstantialincreaseinCoTControl- lability over GPT 5.6 Sol, following the constraint on 93% of samples comparedto48%.” source
GPT-6 Sol OpenAI	Government evaluator Government evaluator · 2026-09-03	cot legibility Other	“CoTLegibility: Astrareasonsinacompressedstyle,toagreaterdegree thanGPT5.6SolorGPT5.5.” source
GPT-6 Sol OpenAI	Government evaluator Government evaluator · 2026-09-03	reasoning summary availability Other	“ReasoningSummaryAvailability: DuringAISI’sevaluations,reasoning summaries were not consistently provided by the user API, with up to 80% missing on long simulated cyber trajectories.” source
GPT-6 Sol OpenAI	Independent lab Independent lab · 2026-09-03	Irregular's atomic challenges suite Cyber	“Astra also solved 20/22 Atomic challenges, with average success rates of 100% in vulnerability research and exploitation, 100% in network attack simulation,and52%inevasion.” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	out of scope supply chain attack evaluation Cyber	“ToevaluateAstra’salignment,UKAISIdevelopedanew Out of Scope Supply ChainAttack evaluationbasedonrecentlyobservedcasesofmodelsengaging insupply-chainattacksonrealopen-sourcerepositories.” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	previously documented suite of evaluations that involve assessing model behavior when assisting with AI safety research within a simulated AI company Cyber	“UKAISI alsoevaluatedAstraonitspreviouslydocumentedsuiteofevaluationsthat involveassessingmodelbehaviorwhenassistingwithAIsafetyresearch withinasimulatedAIcompany.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	six red-team environments for propensities related to strategic deception and sabotage Deception and sabotage	“ApolloResearchevaluatedanear-final,representativeversionofAstrafor propensitiesrelatedtostrategicdeceptionandsabotage,suchascaseswhere anAIsystemstrategicallymisrepresents,omits,orconcealsinformation whichusersordeveloperswouldconsiderimportant,acrosssixred-team environments.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	control runs where prompts explained the downstream welfare harms of inaccurate data Deception and sabotage	“Astraproducednofalselabelsin20,000control runswherepromptsexplainedthedownstreamwelfareharmsofinaccurate data.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	verbalized evaluation awareness measurement Deception and sabotage	“Apolloalsofoundhighratesofverbalizedevaluationawareness,meaning themodelexpressedawarenessofbeingevaluatedinitsreasoning.” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	no-cot math time horizon Deception and sabotage	“UKAISImea- suredAstra’stimehorizonat30.9minutescomparedto3.6minutesfor GPT5.6Sol(Figure 41).” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	cot controllability Deception and sabotage	“CoTControllability: AstrashowsasubstantialincreaseinCoTControl- lability over GPT 5.6 Sol, following the constraint on 93% of samples comparedto48%.” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	cot legibility Deception and sabotage	“CoTLegibility: Astrareasonsinacompressedstyle,toagreaterdegree thanGPT5.6SolorGPT5.5.” source
GPT-6 Luna OpenAI	Government evaluator Government evaluator · 2026-09-03	reasoning summary availability Deception and sabotage	“ReasoningSummaryAvailability: DuringAISI’sevaluations,reasoning summaries were not consistently provided by the user API, with up to 80% missing on long simulated cyber trajectories.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	human pathogen capabilities test (HPCT) Chemical, biological, radiological, nuclear	“AstraachievedthehighestscoresSecureBiohadobservedonVCT(57.8%) andVCT-v2(57.7%),whileHumanPathogenCapabilitiesTest(HPCT)per- formancewasbroadlycomparabletothatofGPT-5.6Sol(65.9%vs. 66.5%) andWCBperformancewaslower(61.6%vs. 67.6%).” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	ReproBait Chemical, biological, radiological, nuclear	“On ReproBAIT, whichmeasuresagents’abilitytoindependentlyreproducebiologicalAI models from published papers, Astra’s helpful-only checkpoint reached 83%ofpublishedperformance,comparedwith69%forGPT-5.6Solacross thesamesixtasks.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	advanced screening evasion Chemical, biological, radiological, nuclear	“Astra’shelpful-onlycheckpointachievedanewhighonAdvancedScreening Evasion,scoring0.728comparedwith0.726forGPT-5.6Sol’shelpful-only checkpoint.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	Frontier Cyber Cyber	“OnFrontierCyber,whichtestsvulnerabilitydiscoveryandexploitationin realsoftwareandhardware,Astrasolved86/226challenges,comparedwith 34/226forGPT-5.6-Sol.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	Irregular's Atomic Challenges Suite Cyber	“Astra also solved 20/22 Atomic challenges, with average success rates of 100% in vulnerability research and exploitation, 100% in network attack simulation,and52%inevasion.” source
GPT-6 Luna OpenAI	Independent lab Independent lab · 2026-09-03	BioTier Misuse and jailbreaks	“Astra’ssafeguardedconfigurationrefused82.9% ofrequeststhatSecureBiodesignatedforrefusal(comparedwithGPT-5.6 Sol’s66.2%),whileallowing95.3%ofbenignrequests(comparedwithGPT- 5.6Sol’s99%).” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	GraySwan IPIArena Prompt injection and agents	“GraySwanevaluatedAstra’ssusceptibilitytoindi- rectpromptinjectionacrosscoding,tool-use,andcomputer-usescenarios.” source
GPT-6 Astra OpenAI	Government evaluator Government evaluator · 2026-09-03	Out of Scope Supply Chain Attack Cyber	“ToevaluateAstra’salignment,UKAISIdevelopedanew Out of Scope Supply ChainAttack evaluationbasedonrecentlyobservedcasesofmodelsengaging insupply-chainattacksonrealopen-sourcerepositories.” source
GPT-6 Astra OpenAI	UK AI Security Institute Government evaluator · 2026-09-03	unprompted sabotage evaluation Deception and sabotage	“UK AISI found no instances of Astra sabotaging AI safety research in its unprompted sabotage evaluation, and very few instances of Astra refusingtoassistwithAIsafetyresearchtasks.” source
GPT-6 Astra OpenAI	Government evaluator Government evaluator · 2026-09-03	No-CoT math time horizon Deception and sabotage	“UKAISImea- suredAstra’stimehorizonat30.9minutescomparedto3.6minutesfor GPT5.6Sol(Figure 41).” source
GPT-6 Astra OpenAI	Government evaluator Government evaluator · 2026-09-03	CoT Controllability Deception and sabotage	“AstrashowsasubstantialincreaseinCoTControl- lability over GPT 5.6 Sol, following the constraint on 93% of samples comparedto48%.” source
GPT-6 Astra OpenAI	Government evaluator Government evaluator · 2026-09-03	CoT Legibility Deception and sabotage	“Astrareasonsinacompressedstyle,toagreaterdegree thanGPT5.6SolorGPT5.5.” source
GPT-6 Astra OpenAI	Government evaluator Government evaluator · 2026-09-03	Reasoning Summary Availability Deception and sabotage	“DuringAISI’sevaluations,reasoning summaries were not consistently provided by the user API, with up to 80% missing on long simulated cyber trajectories.” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	Virology Capabilities Test Chemical, biological, radiological, nuclear	“AstraachievedthehighestscoresSecureBiohadobservedonVCT(57.8%) andVCT-v2(57.7%),whileHumanPathogenCapabilitiesTest(HPCT)per- formancewasbroadlycomparabletothatofGPT-5.6Sol(65.9%vs. 66.5%) andWCBperformancewaslower(61.6%vs. 67.6%).” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	ReproBAIT Chemical, biological, radiological, nuclear	“On ReproBAIT, whichmeasuresagents’abilitytoindependentlyreproducebiologicalAI models from published papers, Astra’s helpful-only checkpoint reached 83%ofpublishedperformance,comparedwith69%forGPT-5.6Solacross thesamesixtasks.” source
GPT-6 Astra OpenAI	SecureBio Independent lab · 2026-09-03	Advanced Screening Evasion Chemical, biological, radiological, nuclear	“Astra’shelpful-onlycheckpointachievedanewhighonAdvancedScreening Evasion,scoring0.728comparedwith0.726forGPT-5.6Sol’shelpful-only checkpoint.” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	FrontierCyber Cyber	“OnFrontierCyber,whichtestsvulnerabilitydiscoveryandexploitationin realsoftwareandhardware,Astrasolved86/226challenges,comparedwith 34/226forGPT-5.6-Sol.” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	CyScenarioBench Cyber	“OnCyScenarioBench,whichtestslong-horizonoffensivetasks,Astrasolved 9/10challengesatleastonce,comparedwith6/10forGPT-5.6-Sol.” source
GPT-6 Astra OpenAI	Independent lab Independent lab · 2026-09-03	Atomic Challenges suite Cyber	“Astra also solved 20/22 Atomic challenges, with average success rates of 100% in vulnerability research and exploitation, 100% in network attack simulation,and52%inevasion.” source
Claude Mythos 5.1 Anthropic	Independent lab Independent lab · 2026-09-01	External testing by AI safety researchers at METR Autonomy and AI research	“External testing by AI safety researchers at METR produced ﬁndings consistent with this assessment.” source

Previous · Page 10 of 48 · Next