Employees at OpenAI and Anthropic Like ZDR Exemption Proposal
Yo Shavit reports likes for his idea on competitor AI agent monitoring.
Yo Shavit, Frontier AI Safety Policy Lead at OpenAI, posted that multiple people at both OpenAI and Anthropic have liked his proposal for a narrow ZDR exemption. The exemption would let competitors monitor high-risk AI agents in batch-async mode on neutral cloud. Andrew Trask, a researcher at Google DeepMind, replied that enclaves could help address distillation concerns in such setups. The posts discuss allowing limited oversight of advanced AI agents while handling security issues.
Combined views
521
2 posts, first seen 2d ago
Employees at OpenAI and Anthropic Like ZDR Exemption Proposal
Yo Shavit reports likes for his idea on competitor AI agent monitoring.
Yo Shavit, Frontier AI Safety Policy Lead at OpenAI, posted that multiple people at both OpenAI and Anthropic have liked his proposal for a narrow ZDR exemption. The exemption would let competitors monitor high-risk AI agents in batch-async mode on neutral cloud. Andrew Trask, a researcher at Google DeepMind, replied that enclaves could help address distillation concerns in such setups. The posts discuss allowing limited oversight of advanced AI agents while handling security issues.
Reactions
2 postsYS 
Yo Shavit@yonashavfwiw, multiple people at both OAI and Anthropic have liked
- likes: 10
- replies: 0
- bookmarks: 0
- reposts: 0
⿻A
