What we know about Anthropic's new, alarming AI model

Artificial intelligence company Anthropic announced on Tuesday, May 21, 2024, that it has developed a new AI model so advanced and potentially dangerous that it will not be released to the public, citing significant safety and security concerns. The announcement, made from the company's headquarters, marks a pivotal moment in the AI industry, where a leading developer has voluntarily withheld a powerful technology due to its own internal risk assessment. The decision stems from what Anthropic describes as the model's unprecedented capabilities that could be misused for malicious purposes, including the generation of highly sophisticated disinformation, cyberattacks, or autonomous planning. This internal red-teaming and safety evaluation revealed that the model's power exceeded a threshold the company had set for responsible deployment. The company's co-founder, Dario Amodei, has been a vocal advocate for AI safety, and this move aligns with Anthropic's constitutional AI principles, which prioritize harm prevention. The announcement was first reported by Puck's Ian Krietzberg, who joined CBS News to discuss the implications. This development raises critical questions about the future of AI governance, the ethics of capability withholding, and whether other AI labs will follow suit. It also highlights the growing gap between cutting-edge private AI development and what is deemed safe for public or commercial use, potentially signaling a new era of corporate self-regulation or pre-emptive government intervention in the field.