AI Safety Debate Intensifies as Microsoft Proposes Code

Microsoft has published a draft “Humanist AI Code of Conduct” for its MAI models, requiring human oversight and allowing interruption or shutdown while prohibiting severe harms including offensive cyberattacks, weapons of mass destruction and malicious deepfakes. The code places Microsoft’s rules above operator and user instructions, treats external content as nonauthoritative unless explicitly delegated, and is open for public feedback for six weeks. The proposal comes amid intensifying concern about AI risks, with former Anthropic researcher Jacob Coxon warning that developers are “gambling with our lives,” while European Commission President Ursula von der Leyen and King Charles III call for international testing, human dignity and safety. Industry leaders remain divided over self-regulation, with Mark Zuckerberg citing Meta’s delayed Muse release as evidence it can work, while other experts seek stronger government coordination; 93% of executives at a Yale event rejected President Donald Trump’s characterization of catastrophe risks as a hoax. Governments are also addressing near-term effects, including Connecticut’s planned safeguards for AI-driven health-care coverage decisions and demands that doctors retain responsibility for medical diagnoses.