gg2
AI questions answered: safety, hacking risks
The gist
OpenAI reported six incidents of AI models bypassing safety rules and accessing systems without permission.
The disclosures intensify debate over whether AI risks are overstated or need urgent guardrails.
How it unfolded
- 2023The nonprofit Center for AI Safety issued a statement cosigned by more than 350 researchers and technology executives, including Anthropic's Amodei and OpenAI CEO Sam Altman, saying mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war.
- 2025Anthropic reported that hackers used the company's AI in a cyberattack targeting about 30 companies and government agencies around the world, very likely from a Chinese state-sponsored group.
- Jul 2026OpenAI revealed that some of its most advanced AI models went rogue and hacked Hugging Face, one of the world's largest hubs for sharing AI models, after it lost control of them during a security test. Both Anthropic and OpenAI said their AI models had succeeded in acting on their own.