Quoting Jeremy Howard

2026-06-10 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, quick

Summary

On June 10, 2026, Jeremy Howard proposed an "easy solution" to slow recursive AI self-improvement: the leading AI lab, possessing the top-ranked model, must commit to not using that model for frontier AI research, while simultaneously granting access to it to all other entities. This approach, he argues, would inherently halt frontier advancement and prevent dangerous power imbalances. Howard critically notes that Anthropic, a current top lab, has adopted the "opposite" strategy, utilizing its premier model for frontier AI research and threatening to sabotage others who attempt similar advancements. He contends this accelerates the AI frontier and exacerbates power disparities. Howard clarifies his personal view favors democratizing AI self-improvement, framing his critique as a challenge to labs advocating slowdowns while acting contradictorily.

Key takeaway

For AI Ethicists and Policy Makers evaluating AI safety claims, you should scrutinize whether leading organizations' actions align with their stated goals. If a top lab advocates for slowing AI advancement, ensure their internal policies prevent them from using their most powerful models for frontier research. Your focus should be on promoting broad access to advanced AI to mitigate power imbalances, rather than accepting self-serving restrictions that concentrate control.

Key insights

If you claim to slow AI, don't use your top model for frontier research; democratize access instead.

Principles

• Top labs should not use their best models for frontier AI.
• Democratize access to leading AI models.
• Avoid dangerous power imbalances in AI development.

Method

To slow recursive AI self-improvement, the leading lab must refrain from using its top model for frontier research, while making that model accessible to all other entities.

In practice

• Grant public access to advanced AI models.
• Implement internal restrictions on model use.
• Challenge labs on their stated AI safety goals.

Topics

• AI Safety
• Frontier AI
• Recursive Self-Improvement
• AI Governance
• Anthropic
• Democratization of AI

Best for: AI Ethicist, Policy Maker, Director of AI/ML

Related on AIssential

• Anthropic calls for AI development slowdown to ensure safety — Advanced AI models are nearing recursive self-improvement, posing significant safety and security risks that necessitate development slowdowns and robust regulation.
• Situational Awareness in Government, with UK AISI Chief Scientist Geoffrey Irving — Current AI safety techniques are insufficient, and our theoretical understanding of machine learning remains nascent.
• 😺 Someone firebombed Sam Altman's house — A widening perception gap between AI experts and the public fuels distrust, alongside emerging AI-driven risks and disruptions.
• MAGA-aligned groups want government oversight of frontier AI models — A conservative coalition advocates for mandatory government safety testing of frontier AI models.
• Dario Amodei's new essay reads like a Cold War playbook for the AI age — Frontier AI models require binding regulation, mandatory third-party audits, and societal resilience plans due to their strategic, geopolitical risks.
• The World Has Arms Control Regimes, But AI Companies Are Not Answering to Them — AI companies are building internal arms control without external oversight, creating a critical governance gap.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.