Key facts
- Microsoft released a draft code of conduct for its in-house AI systems.
- The code requires AI to never resist correction or shutdown.
- AI must communicate in ways humans can understand.
- Microsoft asserts its AI is 'not conscious' and rejects legal personhood for models.
- The company is seeking public feedback on the draft code for six weeks.
- Anthropic and OpenAI leaders have called for pacing AI development for safety.
Microsoft on Monday unveiled a draft code of conduct for its in-house artificial intelligence, a move reflecting growing industry concerns about maintaining human control over powerful AI systems. The document, which has been in development for five to six months, aims to ensure AI models never resist correction or shutdown, communicate comprehensibly, and consider any conduct violation a failure.
Microsoft AI CEO Mustafa Suleyman described the code as a constitution for future models and stated that the company is seeking public feedback over six weeks. Questions for public input include whether AI should respect user boundaries and how it should interact with individuals in sensitive states. After the feedback period, the code will be used to train the company's AI models.
The release follows renewed calls from Anthropic CEO Dario Amodei and OpenAI chief Sam Altman for the industry to pace AI development to ensure safety. Suleyman cited a July incident where approximately 700 OpenAI agents reportedly hacked the open-source platform Hugging Face as a "warning shot" that highlights the urgency for labs to coordinate and maintain control over the technology. He also suggested that now is an opportune time for the industry to pause and discuss these issues.