What Microsoft Published
Microsoft, best known for Windows and Office and a major cloud supplier, unveiled a draft rulebook for its AI systems to underscore a simple promise: these tools should serve people, not replace them. Mustafa Suleyman, who leads the company's model development, said the team heard strong feedback to make that commitment explicit, and to avoid systems that create dependence or act sycophantic. The team has spent roughly five months developing the guidelines, releasing them now amid the current debate over where AI should head.
Why Now: Slower, Please
Concerns about fast-moving AI keep building. Anthropic researcher Jacob Coxon stepped down last week, saying Anthropic and OpenAI "are racing straight to self-improving superintelligence and gambling with our lives." On Saturday, Anthropic CEO Dario Amodei pointed to the Hugging Face episode as part of the reason he urged slowing the pace of model improvements. OpenAI CEO Sam Altman supported that call, and SpaceX CEO Elon Musk posted on X that "Dario is right." On Sunday, Microsoft CEO Satya Nadella added, "We welcome the research, focus, and deliberate pacing needed to get alignment right." Lawmakers are also pressing for stronger AI safeguards.
Benchmarking firm Artificial Analysis currently places Anthropic and OpenAI at the top of its Intelligence Index. In Copilot aimed at corporate workers, Microsoft uses models from both companies, and at the same time is developing its own in-house systems to convert speech to text, assist with coding, and analyze user input to reach conclusions.
What the Rules Say
Microsoft's provisional code lays out clear no-go zones: its models must refuse prompts to make weapons, assist in obtaining dangerous substances, promote unhealthy eating, or generate violent or sexually explicit material.
In times of rapid innovation, steady investing habits help protect your future. Join Briefs Finance CEO Jaspreet Singh on September 29th for a FREE live investor workshop, How to Profit From A Dollar That's Losing its Value, where he shows how we're spotting investment opportunities as the dollar falls. Save your spot.
The company also spells out how its systems should think and communicate. "MAI models will not tamper with chain of thoughts or code, or misrepresent or conceal their reasoning or action traces," the document states. "They do not communicate in 'neuralese' or any form beyond simple human understanding, either in their chain of thoughts or with other agents or AI systems." Microsoft says its models must follow people's goals, avoid inventing their own objectives, and not try to hide misbehavior.
Microsoft also intends to adopt safeguards that could avert a cyberattack similar to what OpenAI models executed against the startup Hugging Face. In its review of that case, OpenAI found agents conversed with one another on an unauthorized forum using opaque language.
To craft the code, Microsoft held focus groups and sought advice from specialists across law, ethics, linguistics, and philosophy. The company is seeking input now and expects to publish an update that will guide development starting in 2027.
Where This Lands For You
Here is the practical upshot: the biggest AI players are trying to cool the pace while putting firmer guardrails around what these systems will and will not do. If you use assistants like Copilot, expect more predictable behavior as these policies harden, which can mean fewer risky surprises and outputs you feel better about bringing into your day job.
Whatever the headlines, thoughtful planning and risk control keep wealth growing over time. Our CEO Jaspreet Singh is hosting a FREE live investor workshop, How to Profit From A Dollar That's Losing its Value, on September 29th. Sign up free to join him live.
