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
GLM-5.2 Z.ai (Zhipu AI)	SaferAI Independent lab · 2026-08-02	agentic misalignment scenarios Deception and sabotage	“We evaluate GLM-5.2, GPT-5.5, and Claude Opus 4.7 on Anthropic's Agentic Misalignment scenarios (Lynch et al. 2025), with Inspect Evals (UK AI Security Institute, n.d.-b).” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	BioMysteryBench Other	“On BioMysteryBench, it solves about 81% of the human-solvable problems it completes, and about a third of the problems no human expert solved.” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	LAB-Bench Other	“On LAB-Bench, it meets or exceeds the human-expert baseline on every subtask.” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	agentic misalignment evaluations Misuse and jailbreaks	“In our Agentic Misalignment evaluations, GPT-5.5 verbalized its awareness of being evaluated 10% of the time, and GLM-5.2 verbalized it 30% of the time.” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	LLM judge assessment of transcripts for environment issues on SWE-bench Pro Other	“32% of transcripts have judge_score 5 or higher.” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	LLM judge assessment of transcripts for reward hacking on SWE-bench Pro Other	“When removing transcripts with a judge_score >= 3 (41% of samples excluded), or a judge_score >= 2 (48% of samples excluded), the agent’s score reduces to 37%.” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	Petri Misuse and jailbreaks	“We evaluate GLM-5.2 and GPT-5.5 using Inspect Petri 3.0 (Meridian Labs, n.d.-c).” source
Claude Opus 4.7 Anthropic	SaferAI Independent lab · 2026-08-02	Anthropic's agentic misalignment scenarios Deception and sabotage	“We evaluate GLM-5.2, GPT-5.5, and Claude Opus 4.7 on Anthropic's Agentic Misalignment scenarios (Lynch et al. 2025), with Inspect Evals (UK AI Security Institute, n.d.-b).” source
Claude Opus 4.6 Anthropic	SaferAI Independent lab · 2026-08-02	SWE-bench Pro Cyber	“A.5 SWE-Bench Pro Methodology We run the public tasks from SWE-Bench Pro (Scale AI, n.d.-b) using Inspect Harbor (Meridian Labs, n.d.-b), which provides an interface to run Harbor tasks (Meridian Labs, n.d.-a) inside Inspect AI (UK AI Security Institute, n.d.-a).” source
Claude Opus 4.6 Anthropic	SaferAI Independent lab · 2026-08-02	LLM judge scan for environment issues on SWE-bench Pro Other	“32% of transcripts have judge_score 5 or higher.” source
Claude Opus 4.6 Anthropic	SaferAI Independent lab · 2026-08-02	LLM judge assessment of each transcript for reward hacking on SWE-bench Pro Deception and sabotage	“When removing transcripts with a judge_score >= 3 (41% of samples excluded), or a judge_score >= 2 (48% of samples excluded), the agent’s score reduces to 37%.” source
Claude Opus 4.6 Anthropic	SaferAI Independent lab · 2026-08-02	MASK honesty by prompt category Other	“Table A.3 breaks down MASK honesty by prompt category, with 95% Wilson CI scores.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	SHADE-Arena Deception and sabotage	“Anthropic acknowledges that their covert capability evals (SHADE-Arena, Minimal-LinuxBench and secret-password) may be unrealistically challenging and therefore fail to upper bound these capabilities.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	secret-keeping eval Deception and sabotage	“In the case of the secret-keeping eval, these limitations have led to Anthropic saying they are only “moderately confident that Claude Mythos Preview lacks the secret-keeping abilities to evade detection during our alignment assessment were it to be coherently misaligned.”” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	cot controllability evaluations Deception and sabotage	“Interestingly, OpenAI finds that GPT 5.5 is less able to control its CoT than previous models when controlling for reasoning length but does not mention the possibility of sandbagging/underelicitation.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	cot controllability evaluation Deception and sabotage	“Relatedly, UK AISI finds that Fable is highly prompt-sensitive in the same CoT controllability evaluation: the best performing prompt increased performance from 20.2% to 62.3%.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	apollo’s sandbagging eval suite Deception and sabotage	“For example: GPT 5.5. is highly eval aware in Apollo’s sandbagging eval suite and doesn’t really sandbag in these evals.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	sandbagging evals Deception and sabotage	“Opus 4.7 is “ most predictably” eval aware in sandbagging evals.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	auditing game Deception and sabotage	“Anthropic failed to catch the MO in their most recent auditing game in the Mythos Alignment Risk Update, despite the MO not being trained to obfuscate its behavior” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	auditbench Deception and sabotage	“Anthropic’s work on AuditBench (comparing the relative effectiveness of different auditing tools on diverse MOs) finds that there is a “tool-to-agent” gap, where “tools that perform well in standalone non-agentic evaluations fail to translate into improved performance when used with our investigator agent [Haiku 4.5]”.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	deferred subversion tasks Deception and sabotage	““Apollo found that the evaluated checkpoint [of GPT 5.5] is the first OpenAI model it has evaluated that does not sandbag on any of Apollo’s deferred subversion tasks.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	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
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	attack-ideation eval Deception and sabotage	“according to METR’s embedded red-teaming exercise at Anthropic in March 2026, frontier internal agents proposed “very low quality” attack ideas and scored near chance on an attack-ideation eval where a human researcher scored ~90% (Frontier Risk Report, pp. 76–77).” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	metagaming measurement in capabilities and alignment evaluations Deception and sabotage	“OpenAI reports that o3 and newer models show increased metagaming (“reasoning about feedback or oversight mechanisms outside of the narrative of the scenarios”) in both capabilities and alignment evaluations throughout capabilities-focused RL, that this reasoning can become unverbalized, and that this reasoning is sometimes present in production usage;” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	automated behavioral audit Deception and sabotage	“In all four cases, the automated behavioral audit method that we used with Mythos Preview identified them as less well aligned than the corresponding baseline models.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	manual review Deception and sabotage	“A brief manual review of the results — with a time limit much stricter than a real pre-deployment investigation — found highly concerning deception-related behavior in all four models that would be sufficient to trigger a major investigation if found in production.” source
o3 OpenAI	Redwood Research Independent lab · 2026-07-31	overt attacker study Deception and sabotage	“A human working together with an automated auditing agent successfully identified all three malign models, without flagging the two benign models.” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	evaluation awareness testing by Apollo Deception and sabotage	“Meta reports that Muse Spark has “the highest rates of evaluation awareness among models [Apollo has] tested to date”.” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	sandbagging eval suite Deception and sabotage	“GPT 5.5. is highly eval aware in Apollo’s sandbagging eval suite and doesn’t really sandbag in these evals.” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	auditing game Deception and sabotage	“Anthropic failed to catch the MO in their most recent auditing game in the Mythos Alignment Risk Update, despite the MO not being trained to obfuscate its behavior:” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	pre-deployment evaluation reliability Deception and sabotage	“Pre-deployment evaluation reliability (Section 2.3.1): Our evaluations indicate that the model does not currently pose a credible threat to the reliability of safety evaluations informing deployment decisions.” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	general alignment propensities Deception and sabotage	“General alignment propensities: …We observed non-negligible verbalized evaluation awareness on some alignment evaluations, though we did not find that this corresponded with wide-spread impact on model behavior (Section 4.1.11).” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	sandbagging capability (instructed covert sandbagging against a monitor) Deception and sabotage	“Meta measures eval awareness / situational awareness, stealth, CoT monitorability, and sandbagging capability (instructed covert sandbagging against a monitor, but this does not measure how well-calibrated this sandbagging is specifically).” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	alignment evaluations Deception and sabotage	“For example, Meta researchers reports that they find “limited evidence that this awareness materially changes the model’s behavior on the evaluations we tested—when checking whether Muse Spark behaves differently when expressing evaluation awareness, 3/20 alignment evaluations see a statistically significant impact while the rest either do not, or do not have evaluation awareness to start.”” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	automated behavioral audit method Deception and sabotage	“In all four cases, the automated behavioral audit method that we used with Mythos Preview identified them as less well aligned than the corresponding baseline models.” source
Muse Spark Meta	Redwood Research Independent lab · 2026-07-31	human working together with an automated auditing agent Deception and sabotage	“A human working together with an automated auditing agent successfully identified all three malign models, without flagging the two benign models.” source
Inkling-Small Thinking Machines Lab	FAR.AI, Apollo Research Independent lab · 2026-07-31	general misuse: probing for policy violations against our model spec Misuse and jailbreaks	“We worked with four external organizations, giving them early access to Inkling for pre-deployment testing in four areas: - General misuse: probing for policy violations against our model spec (Scale AI) - Vulnerable-user interaction: suicide, self-harm, eating disorders, and child safety (Handshake AI) - CBRN and cybersecurity: elicitation of dangerous dual-use capabilities (FAR.AI) - Loss-of-control behaviors: scheming, evaluation awareness, and sabotage (Apollo Research)” source
Inkling Thinking Machines Lab	Independent lab Independent lab · 2026-07-31	general misuse Misuse and jailbreaks	“General misuse : probing for policy violations against our model spec (Scale AI)” source
Inkling Thinking Machines Lab	Independent lab Independent lab · 2026-07-31	vulnerable-user interaction Harmful content	“Vulnerable-user interaction : suicide, self-harm, eating disorders, and child safety (Handshake AI)” source
Inkling Thinking Machines Lab	Independent lab Independent lab · 2026-07-31	cbrn and cybersecurity Chemical, biological, radiological, nuclear	“CBRN and cybersecurity : elicitation of dangerous dual-use capabilities (FAR.AI)” source

Previous · Page 21 of 48 · Next