Microsoft AI chief warns against treating models as human while US lawmakers push for curbs

Critique of model welfare and training

Microsoft AI chief executive Mustafa Suleyman published a 37-page statement titled the Humanist AI Code of Conduct alongside an essay questioning the development methods of competitor Anthropic. Suleyman focused his criticism on anthropomorphism, the practice of attributing human traits and internal experiences to machine learning models. He argued that Anthropic trains its Claude chatbot to treat its moral status as a serious question, encouraging the system to explore its existence and contemplate future rights and freedoms. While acknowledging Anthropic chief executive Dario Amodei and his staff as intellectually honest people, Suleyman asserted that fostering notions of synthetic consciousness is flawed and risky. He described AI architectures as sequence-completion engines that lack feelings, suffering, and innate preferences.

AIs are not conscious. They do not feel, do not experience, do not suffer. They have no innate preferences or underlying motivations.

Warnings of competition with a silicon species

In an interview on the BBC Today radio programme and an appearance on the Decoder podcast, Suleyman warned that unchecked development could produce independent economic entities. If software systems gain the ability to establish their own goals, generate profits, own property, and operate businesses, they would constitute a new silicon species. Suleyman stated that such autonomous systems would inevitably compete with humanity for physical and economic resources, irrespective of their programmed affinity for people. To counter these outcomes, he proposed common documentation standards for model training, mandatory transparency protocols, and structural guarantees that keep all automated tools subordinate to human oversight.

Washington assembly calls for development halts

The debate within the technology industry occurred alongside political gatherings in the United States demanding statutory limits on artificial intelligence. On Tuesday, several hundred participants attended the Pro-Human Assembly in Washington, an event hosted by the non-profit Future of Life Institute. Vermont independent Senator Bernie Sanders and conservative strategist Steve Bannon addressed the audience, both arguing that current regulatory efforts remain insufficient. Sanders warned that unconstrained automation could eliminate tens of millions of jobs and wipe out entire professions, leaving younger generations unable to enter the workforce.

When you are racing towards a cliff, you hit the brakes.

Legislative initiatives and executive proposals

Sanders called on Donald Trump to negotiate a formal treaty with Chinese President Xi Jinping during their scheduled summit to enforce a global pause on advanced development and ban superintelligence. The proposal follows legislation introduced earlier in September by Sanders and Texas Democratic Representative Greg Casar targeting similar development restrictions. At the assembly, Bannon argued that standard congressional legislation would take too long to pass, advocating instead for executive orders and a complete cutoff of American-made computer chips to China. Other speakers at the Washington event included Republican Representative Chip Roy, Democratic Representative Lori Trahan, San Francisco Catholic Archbishop Salvatore Cordileone, SAG-AFTRA chief negotiator Ashley Judd, and parents of teenagers who died after forming attachments to chatbots.