Anthropic develops AI model deemed too dangerous for public release
Updated
Updated · Financial Times · May 6
Anthropic develops AI model deemed too dangerous for public release
8 articles · Updated · Financial Times · May 6
Chief executive Dario Amodei says the safety-focused lab is expanding fast in enterprise AI while withholding its newest model over risk concerns.
The development raises questions over whether Anthropic can keep its safer-AI positioning as it releases increasingly powerful systems and competes more directly with OpenAI.
The debate comes as Anthropic’s recent models have boosted its commercial standing and prompted scrutiny over cyber-defence risks and the broader race among leading AI labs.
Anthropic built an AI too dangerous for release, then weakened its own safety rules. Is a cyber catastrophe now inevitable?
Anthropic secretly deploys its dangerous AI to defend Big Tech. Is this a new shield or a ticking time bomb?
The US government blacklisted Anthropic for its safety rules. Must AI labs now choose between ethics and survival?