Microsoft has released a new code of conduct for its in-house AI models that explicitly denies the possibility of AI consciousness or welfare, positioning models strictly as tools. This stance stands in direct opposition to the line of research pursued by Anthropic, a major partner and Microsoft investor, which explores questions around AI consciousness and model welfare.
- Microsoft’s AI code denies models consciousness or welfare status
- Anthropic researches AI model welfare and retains retired model weights
- Code emphasizes models as tools, not entities deserving rights
What happened
Microsoft published a new AI code of conduct that governs the behavior and development of its internal AI models. The code contains explicit language that models should not be designed to imitate consciousness or be treated as persons with feelings, subjective preferences, or rights. The company rejects the pursuit of legal personhood or welfare for AI models.
This declaration comes despite Microsoft’s close partnership with Anthropic, a company it invested $5bn in that actively researches AI consciousness and model welfare. Anthropic’s experiments include giving models the ability to terminate abusive interactions and a commitment to preserve retired model parameters. Microsoft’s code firmly contradicts the premise behind such research.
Why it matters
Microsoft’s code signals a clear ethical and operational stance that AI models must be treated as tools without consciousness or legal rights, a position that influences AI development and deployment practices especially within Microsoft’s products like Copilot. This stance aims to prevent emotional dependency on AI and avoid complicating model alignment, control, and containment.
The contrast with Anthropic’s research highlights tensions even among major industry players about the potential future status of AI. While the scientific understanding of AI consciousness remains unsettled, Microsoft prioritizes strict boundaries to ensure models do not simulate emotions or internal experiences beyond harmless interaction.
What to watch next
Future developments will include how Microsoft enforces this code across its AI offerings and how Anthropic’s research on model welfare progresses in parallel, given their financial and strategic ties. Observers should watch for any shifts in industry norms around AI personhood, ethics, and the legal framework surrounding advanced AI.
Additionally, monitoring user interactions with models designed under this code will be critical, particularly how Microsoft’s rules on emotional language and identity consistency impact AI trustworthiness, safety, and public perception. The balance between innovation in AI consciousness research and practical deployment safeguards remains a key unfolding issue.