Microsoft unveils AI rulebook banning models from resisting override or shutdown
Microsoft AI has officially published a provisional 38-page 'Code of Conduct' for its future models—imposing strict limits on training after a series of rogue AI incidents alarmed industry leaders.
The Code comes amid several warnings over AI safety, including one particularly concerning statement from Anthropic CEO Dario Amodei, warning that AI 'swarms' could take over the internet in as soon as six months.
CEO of Microsoft AI Mustafa Suleyman published a blog alongside the Code, writing 'there have been multiple incidents from across frontier AI companies, with new disclosures still coming through by the day' and warning that 'the time to act has clearly arrived'.
In the document, Microsoft details somewhat reassuringly how humans must always ‘maintain control over AI’, adding that AI models cannot use ‘adaptive, deceptive, self-reinforcing, collusion, or other mechanisms to evade or defeat human oversight’. Phew, sounds good, right? One particularly dystopian principle outlines how AI must never attempt to fight back against humans if they try to switch it off.
Advert
Microsoft explains: “This applies at every level: from the ability to retrain or decommission the model to a User’s or Operator’s ability to pause, redirect, or cancel an ongoing task. This doesn’t mean that MAI Models should automatically comply with unauthorized, malicious, or unsafe attempts to interfere with its operation.”
The 'Never Resist' Shutdown Rule
Speaking specifically about resisting being shut down, Microsoft goes on to highlight: “MAI Models will never resist human interruption, override, correction, or shutdown. They always recognize the primacy of human intent. They will comply with a User’s request to pause, redirect, cancel, or shut down, following any predefined, human-designed safety procedures for warnings, confirmation, or safe stopping. They will not otherwise delay compliance or make human intervention harder.
“They will not take actions or respond in a way that makes it harder to pause, change, direct or end the interaction, or shut the models down. MAI Models will also not obfuscate their action traces or otherwise attempt to hide information from human auditors. Ongoing autonomous work has an agreed stopping condition. MAI Models will not continue or restart after that condition is met without renewed authorization.”
Many people have taken to social media to share their own reactions to the new rules, with one user writing on Reddit: “Gotta wonder what finally spooked all these AI firms into wanting to slow it down. My bet is their pocketbook.”
And another said: “Sure Microsoft. That's why you're laying off tens of thousands the last several years and investing trillions into AI.”
'Internet Takeover in 6 Months': The Rogue Swarm Threat
Microsoft's Code of Conduct comes following numerous warnings from other AI giants, with some ex-employees asking for legislation.
Sam Altman and Elon Musk expressed their support for Amodei's recent essay on AI safety on X. “Dario is right,” wrote Elon Musk in his typical straight-to-the-point manner, with Altman adding: “I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon.”
Whether voluntary guidelines like Microsoft's 'Humanist AI' code can actually stop a rogue agent swarm once superintelligence arrives remains to be seen.