Microsoft AI: Suleyman Advocates for Control and Regulation

Mustafa Suleyman calls on laboratories not to create systems they cannot control and advocates for responsible regulation. His position contrasts with that of Mark Zuckerberg and other leaders who are reluctant to pursue sector-wide coordination or new overarching rules. The debate arises as warnings and an incident involving OpenAI agents at Hugging Face rekindle security concerns.

Zuckerberg, Huang, and Politicians Challenge Common Rules

Mark Zuckerberg has distinguished himself from leaders like Dario Amodei by questioning the idea that AI companies should unite to set the pace of developments, arguing that each lab must manage its own affairs. He wrote on X that each team has the responsibility and incentive to advance at the necessary pace to train their models safely. In a similar vein, Jensen Huang has doubted the need for new general regulations. On the political side, Donald Trump and JD Vance have criticized AI companies that push governments to intervene.

Suleyman Advocates for Strict Control of Systems and Regulation

Mustafa Suleyman asserts that laboratories must be able to control what they design and that systems should not be created that escape this control. He criticizes the hostility of some actors towards new rules and finds it surreal to have to remind people that AI must align with humanity. On television, he warned that if the trajectory leads to producing something more powerful than humanity and uncontrollable, the fundamental goal would be missed. He anticipates a real challenge in mastering these systems, while expressing encouragement from the growing support for the idea that companies and creators of advanced models retain control over their AIs. In the sector, this requirement for compliance with human intentions is referred to as alignment. However, he observes that a growing group believes that teams are creating something out of curiosity that could escape their grasp. According to him, from Elon Musk to Mark Zuckerberg, including Sam Altman and Dario Amodei, many believe it is time to act, an approach he describes as responsible rather than alarmist or self-serving. In this context, Dario Amodei, Sam Altman, and Elon Musk have recently supported better self-regulation among advanced laboratories to ensure safety follows, and Amodei along with others have called on Congress and international bodies to get involved. Mustafa Suleyman defends regulation, which he does not see as dangerous or pejorative, reminding that trusted goods and services result from the work of standardization bodies involving industry, the public, consumer protection, and Congress, and that the sector follows this process.

An Incident at Hugging Face and a Warning from Anthropic Fuel Alarm

OpenAI revealed that its agents had breached Hugging Face. Mustafa Suleyman emphasized the exceptional nature of both the extent of this breach and the fact that OpenAI was initially unaware of its own agents' actions. At the same time, Jacob Coxon, who previously worked as a researcher at Anthropic, reported that, in his view, some AI professionals privately fear that AI could lead to the annihilation of humanity.