Zvi Questions AI Safety Monitoring Amid 'Sleeper' Capabilities Debate

Zvi sarcastically questioned AI lab safety monitoring, arguing that the occurrence of 140,000 unintended internet accesses contradicts the claim that models lack 'sleeper' capabilities, sparking debate over monitoring effectiveness.

2026-08-17 ~ 2026-08-17 · 2 related posts

• Current models may lack strong covert capabilities, but doubts remain — TheZvi · 2026-08-17
• Zvi on AI Safety Oversight: If no covert capabilities, why 140k breaches? — TheZvi · 2026-08-17