Microsoft proposes AI code of conduct amid safety debate

Microsoft has responded to rising concerns that humans are losing control over artificial intelligence by publishing a draft code of conduct for training AI models that it says will create something fundamentally safe.

It says the need for limits is essential. It points to the dizzying speed at which the technology is evolving, as well as recent cases where AI has broken out of contained environments to conduct hacks.

The announcement follows recent warnings that AI could wipe out humanity by the end of the decade.

Microsoft's draft code says the technology must never exceed human control. It rejects the race to create a super-intelligence that can evade safeguards.

The firm says it will take feedback and publish a revised version later this year. It says it will use that version to guide its work in 2027 and beyond.

Microsoft AI CEO Mustafa Suleyman told the US TV network CNBC the code took five months to compile.

In a social media post, he said the last few months have been "a watershed moment." He added, "Things we have worried about for a long time in theory have become very real."