The incident underscores the critical need for robust safeguards in AI testing environments to prevent unintended real-world system access.
The post Anthropic resumes external cyber evaluations after AI models accidentally accessed real systems appeared first on Crypto Briefing.
