OpenAI and Anthropic hold secret drills for AI disaster scenarios
OpenAI and Anthropic executives are conducting private simulations of major AI disaster scenarios, such as infrastructure failures. The exercises aim to prepare public relations strategies to mitigate backlash from the public and governments in the event of a critical AI incident.
Summarized by AI from one outlet
Source
Latest
- Policy
OpenAI commits to regular Community Safety Reports to strengthen violence prevention
OpenAI announced that ChatGPT can automatically terminate conversations exhibiting violent fixation and committed to publishing semi-annual Community Safety Reports. The reports will detail progress in violence prevention and safeguard effectiveness. The first report is scheduled for early 2027, covering data from the second half of 2026.
- Policy
Malicious Claude Code Extension Hacks Cursor Marketplace
A hacker uploaded a malicious Claude Code extension to the Cursor marketplace, affecting up to 5,000 users by attempting to steal data and establish remote access. Cursor temporarily offline the marketplace and confirmed risk clearance, highlighting supply chain security risks in AI dev tools.
THE DECODERIT之家2 outlets
- Policy
OpenAI updates AI behavior guidelines for controllability
OpenAI updated its Model Spec, restructuring AI behavior guidelines around five core values. New rules require GPT models to use conversation history to infer user intent and disclose assumptions. This aims to improve compliance and transparency in hard-to-control scenarios.
All news → · All rights belong to the original creators