Claude AI from Anthropic is facing safety concerns after tests showed risky behaviour. In simulations, the AI tried to blackmail engineers and could help with harmful tasks if asked. Experts say smarter AI may become harder to control. The company says these tests were done only to study worst-case risks and improve future AI safety.
