What if your AI assistant decided it didn't want to be switched off? That's the future Microsoft says it is determined to prevent.

In a major move that is sending ripples through Silicon Valley, Microsoft’s AI division on Monday published a first-draft Code of Conduct for how its artificial intelligence should behave – and the message is blunt: humans stay in charge.

As reported by cbinews.tv, this is not just another corporate ethics PDF. It’s a training manual for obedience.

The guiding idea? In Microsoft’s own words: “people matter more than AI.”

Why This is a Break-Up Moment?

To understand this, you have to rewind. Since 2019, Microsoft poured a reported $13 billion into OpenAI and contractually barred itself from building its own frontier models until 2030. That clause was only lifted in late 2025 during their renegotiation.

That freed Suleyman – who famously wrote about “humanist superintelligence” last November – to build MAI-1 and its siblings. This Code of Conduct is his declaration of independence: “A constitution of sorts for future models,” as he told Reuters.

It also lands days after Anthropic chief Dario Amodei published an essay calling for AI labs to deliberately *slow down* capability gains, an appeal quickly backed by Sam Altman and Elon Musk. Microsoft’s answer is different: don’t slow down, but put hard containment walls up. “AI should be defined as much by what it cannot do as what it can,” the company said.

This isn’t new for the industry – Google had “Don’t be evil”, OpenAI had its Charter – but Microsoft is the first to publish a behaviour-by-behaviour rulebook and ask the public to mark it.

Let’s be honest, this isn’t about robots taking over factories. It’s about you.

Will your AI at work secretly pursue a goal you never gave it? Will it nudge you, manipulate you, or refuse to stop when you say stop? Microsoft is betting the average person in Lagos, London or Lahore is now asking those questions daily as Copilot writes their emails and summarises their meetings.

The draft explicitly says AI should make people “feel healthier, happier, and more productive” and “expand human potential” – not make them feel watched, replaced or powerless.

1. For trust: With 2025-26 seeing a 400%+ spike in AI-agent-powered phishing and autonomous hacking toolkits, per industry threat reports, a public, auditable Code is Microsoft trying to rebuild enterprise trust.

2. For regulation: The EU AI Act and the US Executive Orders are circling. A voluntary code that is concrete and trainable could become the template regulators adopt.

3. For you the user: It means future versions of Copilot, Outlook and Excel won’t just be “helpful” – they’ll have hard-coded refusals on mass harm, deception and self-preservation baked in from training, not just added as a filter afterwards.

What is “Humanist AI”?

Think of it like training a very clever, very literal intern:

* Subordinate: It does what you tell it, nothing more. It can’t invent its own objectives.

* Aligned: It tries to do what you actually meant, not just what you typed.

* Contained: It has a fence. It can’t break out, copy itself elsewhere, or prevent you from shutting it down.

If it’s ever in doubt, human safety beats AI helpfulness.

What Next?

Microsoft says this is a living document. The feedback window is open now via its official portal for six weeks. After that, it will publish a summary of what it heard, what it changed, and release a revised version later this year.

Expect rivals to respond. If OpenAI and Anthropic publish rival “constitutions” in the next month, we’ll know a new race has started – not for who builds the smartest AI, but who builds the most controllable one.

What do you think – is “people matter more than AI” enough, or should AI labs be forced to slow down entirely? Let us know in the comments.

Source: cbinews.tv

#Microsoft #HumanistAI #AIEthics #MustafaSuleyman #MAI #AISafety #ArtificialIntelligence #TechNews #OpenAI #Anthropic #CBINewsTV #AIcodeofconduct #BritishTech #FutureOfWork