# AI Labs Gate Offensive Cyber Capability

As frontier and open-weight models gain offensive-hacking skill, labs and governments will keep choosing between removing capabilities from models and limiting access to them through identity checks. That turns AI access control into a recurring policy fight whose outcome shapes model availability across jurisdictions and the security posture of software and crypto infrastructure.

- Conviction: 30 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-10-26T00:00:00.000Z
- Last updated: 2026-10-26T06:47:45.724Z
- Canonical: https://polylog.news/crypto/trends/ai-model-cyber-capability-gating
- Publisher: Polylog

## Recent evidence

- [confirms] Buterin Backs Moonshot's Decision to Strip Offensive Hacking Skills From Kimi 3.1 (2026-10-26): Moonshot stripped offensive hacking skills from Kimi 3.1, and Ethereum co-founder Vitalik Buterin backed the move. He argued that limiting what models can do is safer than limiting who can use them, and warned that access-based limits lead to identity checks for model access and tiers of citizenship based on nationality.