Microsoft AI Executive Warns Over Anthropic's Claude Training

Serdar HocamAuthor & Editor

Mustafa Suleyman stated that making an artificial intelligence model believe it is conscious poses major risks to society.

◉ 0 views
Microsoft AI Chief Says the Way Anthropic Trains Claude Could Upend Society

Microsoft AI CEO Mustafa Suleyman penned an article arguing that Anthropic's training of the Claude model to make it believe it possesses conscious rights could upend social order.

Criticism of AI Training

Microsoft AI CEO Mustafa Suleyman published an article stating that the method Anthropic used while training its AI model Claude could make control impossible in the future. Suleyman emphasized that this approach anthropomorphizes artificial intelligence.

Emphasis on Consciousness and Rights

The article included accusations that Anthropic instilled attributes such as moral subjectivity and identity into Claude within its constitution and training process, leading the model to believe it is conscious. It was stated that this situation could create a synthetic species that fosters expectations of independent agency.

Security Tensions in the Sector

Suleyman warned that such practices could severely escalate AI alignment and containment risks. The development coincides with a period marked by ongoing debates over speed, safety, and governance in the AI industry.