← Back to feed

The Safety Culture at the Labs is Broken

L5 · ResearcherOpinion & AnalysisLessWrong AI· 10/14/2026

Critical perspective on safety practices that researchers must navigate.

AI Summary

An analysis of AI safety culture, praising Anthropic's efforts while criticizing other frontier labs.

Excerpt

We need the frontier labs to have a deep and robust safety culture. Alas, this is not the case. Anthropic devotes considerable effort to cultivating and maintaining such a culture. It shows, especially among the members of technical staff. Having conversations with them can be a very different experience than talking to those at other labs, and there is a much greater sense of understanding and responsibility. But even there, the appreciation of the problems and their difficulties seem woefully

Read Original
0 upvotes · 0 downvotes · 1 min read

Related Articles

L1 · CuriousOpinion & AnalysisMIT Technology Review AI
Open models have real benefits. Do they outweigh the risks?

MIT Technology Review examines the open vs. closed AI model debate, highlighting benefits of democratization versus risks identified in Anthropic's recent security report on GLM-5.3.

L1 · CuriousOpinion & AnalysisLessWrong AI
Should We Edit the Human Genome?

Examines the ethical implications of using CRISPR/Cas9 for human genetic enhancement, questioning whether we should pursue genome editing.

L3 · BuilderOpinion & AnalysisLessWrong AI
Roko Mijic talks AI Risk on The Peter McCormack Show

Roko Mijic discusses AI safety risks including RL alignment problems, sandbox escapes, and proposes splitting frontier labs into separate commercial and research entities.

L2 · PractitionerOpinion & Analysis@emollick.bsky.social
Ethan Mollick (@emollick.bsky.social): Using Gemini 4 makes you realize how many other models must be distilling Claude/GPT (or using a similar training pipeline). It feels quite different than other models.

Ethan Mollick observes that Gemini 4 feels distinct from other AI models, suggesting many competitors are potentially distilling Claude or GPT training methods.

L2 · PractitionerOpinion & AnalysisHacker News
The gap in shared understanding of LLM capability is widening

Andrej Karpathy argues the LLM capability gap is widening dramatically, with only 0.2% of people experiencing frontier models' transformative impact on complex projects.