Microsoft has introduced a comprehensive AI code of conduct establishing strict boundaries for advanced model behavior. The new framework prioritizes safety and alignment, enforcing absolute constraints that prohibit models from engaging in cyberattacks, generating deepfakes, or deceiving human operators.
The company aims to address future challenges as systems edge closer to surpassing human performance. Under the new guidelines, Microsoft’s models must reinforce human agency rather than bypass oversight, ensuring they cannot evade shutdown mechanisms or operate entirely beyond human control.
- Microsoft releases a new AI safety and alignment code of conduct
- Absolute bans implemented against cyberattacks and deepfake creation
- Measures designed to prevent AI systems from evading human oversight
- Commitment to maintaining reliable shutdown options for authorized users
Sources:
