How Anthropic Learned Mythos Was Too Dangerous for the Wild
Bloomberg.com • • business
Key Points:
- Nicholas Carlini, a prominent AI researcher, tested Anthropic PBC's new AI model, Mythos, during a wedding in Bali to evaluate its potential risks.
- Anthropic employs Carlini to rigorously stress-test its AI models for vulnerabilities that hackers might exploit for espionage, theft, or sabotage.
- Carlini was surprised by the capabilities of Mythos, indicating the model's significant potential impact and risks.
