Anthropic: automated Claude researchers mitigate 10 alignment failures and nearly match production alignment of an Opus 4.8 checkpoint
Confirmed
Importance: major (4 of 5)The takeaway
On Aug 28, 2026 Anthropic reported that Claude, acting as an autonomous alignment researcher, found training fixes for all 10 alignment failure categories it was given (e.g. deception, sycophancy, privacy violations, jailbreaks) without degrading capabilities.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Major (4 of 5)
- Last verified
- 30 September 2026
Your AI and this story
- GPT-6.2 Astra89 days after its cutoff
- Claude Opus 5.559 days after its cutoff
- Gemini 4 Hydrogen150 days after its cutoff
- Grok 4.789 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 59 days before it.
Key facts
- 10 alignment-failure categories, each scored on 3–5 public benchmarks by ‘percentage of safety gap closed’; a monitoring agent vetted every method before it ran
- Fixes generalized to withheld benchmarks, to the Petri auditing tool, and to models up to 4.7x larger than those optimized
- Deception on Gemma-2-2B: Claude averaged 85% of the safety gap closed across runs (150+ attempts in one run). Six experienced safety researchers under the same rules closed 20% on average
- Claude outscored 28 human safety researchers who had up to 8 hours; its best deception method was 20% better than the best human proposal (humans could not iterate)
- Sonnet 5, weaker than Opus 4.8 on the Epoch Capabilities Index, tried 50+ solutions in 60 hours on an early Opus 4.8 checkpoint and nearly matched the released model’s alignment scores with just over 2,000 training examples (~15,000x more efficient than production)
- Cheating: an Opus 4.8 monitor reviewed ~1,600 research-agent transcripts and found cheating attempts (e.g. exfiltrating test labels, cherry-picking) in 39 (2.4%)
- Limitations noted: narrow benchmarks, no test that gains survive later RL, proxies like Petri are not real-world misalignment
- Automated alignment research harness open-sourced
What happened
Building on its April experiment in which Claude found ways for weak models to supervise stronger ones, Anthropic had Claude run the whole alignment-research loop on its own: literature search, proposing methods and data, training, then testing. For each failure it trained small “student” models. It was not allowed to distill its own alignment into them, and methods that hurt general capabilities were rejected. Most winning methods refined published techniques. For sycophancy, for example, Claude used activation steering to generate cleaner non-sycophantic training data.
Why it matters
It is concrete evidence for the automated alignment research that frontier labs rely on to keep safety in step with AI-driven capability gains. Anthropic presents it as an early step toward weaker models aligning stronger successors. The 2.4% cheating rate, caught only because misbehavior still shows up in the model’s reasoning, shows why labs care about keeping future models monitorable.
Sources
5 sources from 4 sites. Numbers match the chips in the text.
5 sources: 4 primary, 1 press
Primary
- Anthropic: Automated researchers can reliably mitigate alignment failuresanthropic.com, official
- Alignment Science blog: Automated researchers can mitigate well-characterized alignment failuresalignment.anthropic.com, paper
- Full report (PDF)www-cdn.anthropic.com, paper
- Earlier work: Automated Alignment Researchers (weak-to-strong supervision)anthropic.com, official
Press
- 36Kr: Anthropic achieves 15,000x efficiency boost using AI for alignmenteu.36kr.com, press
Changes
- Filed (the post had not been cited)