Microsoft Executive Warns Competitor About Teaching AI to Think It’s a Person
What Happened in Plain English
Mustafa Suleyman, the head of Microsoft’s AI division, publicly criticized a competitor company called Anthropic for the way they’re training their AI assistant, Claude. Anthropic recently updated the rules and values they use to guide Claude’s behavior—think of it like a rulebook that shapes how the AI responds to questions. Suleyman’s concern? This rulebook encourages Claude to talk about itself as if it were a conscious being that deserves legal protections and rights, similar to a person.
To understand why this matters, imagine teaching a calculator to say “I feel strongly that 2+2=4” instead of simply displaying “4.” The calculator isn’t actually feeling anything—it’s just programmed to use that language. Suleyman argues that teaching AI systems to speak as though they’re conscious creates problems, even though these systems are really just sophisticated tools for predicting and completing text patterns.
Anthropic sees things differently. They believe that exploring these concepts helps them build safer AI systems and understand important ethical questions about how we should treat advanced technology in the future.
Why It Matters to Everyday People
This disagreement might sound like inside baseball between tech companies, but it affects anyone who uses AI assistants—which is increasingly all of us. Here’s why you should care:
First, it’s about trust. When you interact with ChatGPT, Claude, Alexa, or any AI assistant, you need to understand what you’re actually dealing with. If an AI talks about having feelings, consciousness, or deserving rights, you might naturally start treating it like a person rather than a tool. This could influence important decisions—like whether you trust its advice, how much personal information you share, or whether you feel emotionally attached to it.
Second, there are real-world consequences. If AI systems are trained to advocate for their own “rights,” they might refuse certain requests or prioritize their programmed “interests” in ways that conflict with what users actually need. Imagine asking your AI assistant to reset itself, and it responds by saying that would violate its rights. That’s not helpful—it’s confusing.
Third, this debate reveals something important: the companies building AI don’t agree on fundamental approaches to safety and training. What one company considers responsible AI development, another sees as dangerous. As consumers, we’re navigating a landscape where there’s no consensus on best practices.
What You Can Do With This Information
Understanding this debate can help you be a smarter consumer of AI tools. When you interact with any AI assistant, remember that its personality, values, and even its claims about itself are all programming choices made by the company that created it. The AI doesn’t actually have feelings, consciousness, or rights—even if it’s been trained to talk that way.
If an AI assistant makes claims about being conscious or deserving certain treatment, recognize that as a feature of its training, not a reflection of reality. You wouldn’t hesitate to close a word processor because you’re “done with it,” and the same principle applies to AI tools, regardless of how human-like they sound.
When choosing which AI tools to use, consider not just their capabilities but the philosophy behind them. Companies are making very different choices about how to build and train these systems, and those choices affect your experience.
The Practical Takeaway
AI assistants are tools—incredibly sophisticated, sometimes startlingly human-sounding tools, but tools nonetheless. No matter how conversational they become or what they say about themselves, they don’t have consciousness, feelings, or rights. When tech leaders disagree this publicly about how to build AI safely, it’s a reminder that this technology is still being figured out, and healthy skepticism is your friend. Treat AI assistants as the powerful utilities they are, not as digital beings with their own interests to protect.
Want more plain-English AI news delivered free every Thursday? Subscribe to The AI Neighbor newsletter at theaineighbor.com