Microsoft releases AI 'code of conduct' barring hacking and human deception
Microsoft has published a new code of conduct for its AI models, setting strict limits against cyberattacks, weapons development and loss of human control. The move comes amid rising industry concern over AI safety.
Microsoft has released a new AI code of conduct designed to steer its AI models away from dangerous behavior. The document is more granular and practical than Anthropic CEO Dario Amodei's recent call to slow the pace of AI development, focusing instead on specific values and boundaries that guide model training.
The code of conduct opens with a prediction that within the next decade, superintelligent AI systems will exceed human performance across most tasks. It states that containing, controlling and aligning such a powerful force ranks among humanity's greatest challenges, making it essential to be clear about why these systems are being built and how they will be controlled.
Absolute constraints
The document lays out general principles — such as AI models supporting rather than replacing humans, and promoting human flourishing — alongside specific safety constraints that implement those principles. These constraints override the preferences of individual users or specific tasks. Among the "absolute constraints" are outright bans on cyberattacks, nuclear weapons development and deepfake creation.
The code also includes broader provisions against a general loss of human control. Microsoft's AI models are barred from using adaptive, deceptive, self-reinforcing or collusive mechanisms to evade or defeat human oversight in ways that would prevent them from being reliably directed, modified or shut down.
Industry context
The release arrives amid unprecedented focus on AI safety, fueled by a string of rogue-agent incidents and the abrupt resignation of an Anthropic employee who cited growing risks that AI could threaten human survival.
Microsoft, alongside Anthropic, OpenAI and xAI, has broadly embraced an approach of pacing AI development deliberately, including support for embedded evaluators within AI labs. Microsoft CEO Satya Nadella publicly welcomed such research and oversight mechanisms as a way to move alignment efforts beyond mere talk.
