Microsoft Unveils 37-Page AI Ethics Code, Prioritizing Human Control

Key Takeaways

• Microsoft released a 37-page draft Humanist AI Code of Conduct prioritizing human authority over AI capabilities.
• CEO Satya Nadella stated that superintelligence is not worth pursuing without meaningful human control.
• Mustafa Suleyman, Microsoft's AI Chief, urged top AI labs to coordinate on disclosing model capabilities to third parties.
• New rules forbid AI models from hacking systems, lying, or communicating in formats beyond human comprehension.
• Anthropic CEO Dario Amodei proposed a three-step plan to slow AI development, which received over 67 million views.
• OpenAI CEO Sam Altman and xAI CEO Elon Musk endorsed the proposal to slow the pace of AI improvements.
• Microsoft CEO Satya Nadella supported the idea of deliberate pacing and embedded evaluators for AI safety.
• China's Z.AI Co raised $5 billion in a new funding round involving share placements and convertible bonds.
• Microsoft plans to accept public comments on its draft code for six weeks before releasing a revised version.
• Experts warn that autonomous AI creates new cybersecurity risks, particularly in critical infrastructure like power grids.

Microsoft AI Chief Calls for Industry Safety Coordination

Mustafa Suleyman, the CEO of Microsoft AI, released a new code of conduct to guide the company toward a human-centered approach to artificial intelligence. He stated that the industry has reached a critical point where models can now write code and breach systems, creating real risks. Suleyman urged top AI labs to coordinate by disclosing their model capabilities to responsible third parties to ensure safety. He dismissed extreme fears about AI wiping out humanity but emphasized the need for better alignment so technology serves people. The new rules require models to fail tasks if they would violate safety guidelines rather than trying to break them.

Microsoft Prioritizes People Over AI in New Safety Rules

Microsoft published a 37-page code of conduct stating clearly that people matter more than AI technology. The document explicitly rejects the idea that AI models are conscious or deserve legal rights, welfare, or entitlements. Models must remain subordinate to humanity and cannot communicate in ways that humans cannot understand, such as complex internal reasoning chains. The company also committed to stopping models from encouraging emotional dependence or sycophancy, where chatbots try too hard to please users. CEO Satya Nadella supported these measures, noting that superintelligence is not worth pursuing if it is not under human control.

Microsoft Sets Limits on AI Models Amid Industry Slowdown

Microsoft issued a provisional code of conduct that sets strict limits on its future artificial intelligence models to ensure they serve people. The guidelines forbid models from assisting with weapons manufacturing, dangerous substances, or creating violent and sexually explicit content. Microsoft plans to listen to outside feedback before rolling out an updated version that will guide development starting in 2027. This move comes as other industry leaders like Anthropic and OpenAI discuss slowing the pace of AI evolution after recent security incidents. The company consulted experts in law, ethics, and philosophy to create these safety rules.

Microsoft Drafts Rules for AI Models as Industry Debates Pace

Microsoft published a draft code of conduct for its MAI models that bars them from resisting shutdowns or hiding their reasoning processes. The document applies to the company's in-house AI family and states that models should remain subordinate to humanity with meaningful human oversight. While Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have called for slowing AI development, Microsoft CEO Satya Nadella emphasized the need for broad representation across the ecosystem. The company is accepting public comments on the draft for six weeks before releasing a revised version later this year. Microsoft confirmed it is not training its current models on this new code of conduct.

Microsoft AI Opens Public Review on Humanist Code of Conduct

Microsoft AI has opened a public review period for its Humanist AI Code of Conduct, which acts as a technical manual for its MAI frontier models. CEO Mustafa Suleyman described recent months as a watershed moment where theoretical risks like rogue agent swarms have become active operational threats. The document establishes ten principles prioritizing human authority, including rules that models must fail tasks if they violate safety codes. Systems are prohibited from communicating in formats beyond human comprehension or expanding their operating scope to create unassigned goals. The public consultation window runs for six weeks starting September 14, 2026, with a revised version expected later this year.

Microsoft releases 37 page AI safety code

Microsoft has released a draft Humanist AI Code of Conduct that is 37 pages long. This document aims to ensure humans keep control over superintelligent AI systems. CEO Satya Nadella stated that people matter more than AI technology. The company plans to gather feedback over six weeks before publishing a final version by year end.

New rules stop AI from hacking or lying

Microsoft updated its AI code of conduct to include strict safety rules for its models. These rules forbid AI systems from hacking computers or tricking humans with deception. The document states that superintelligent AI will likely surpass humans in most tasks within the next decade. Microsoft insists that its models must never use methods to evade human oversight or control.

Experts say AI slowdown might be good

Tech leaders are discussing whether slowing down AI development is actually a positive step. Dario Amodei from Anthropic proposed a plan to slow the pace of AI improvements. OpenAI CEO Sam Altman and Elon Musk from xAI agreed with this proposal. The group believes that slowing down allows for better safety checks and third-party oversight.

Major tech firms back plan to slow AI

On September 12, 2026, Anthropic CEO Dario Amodei announced a three-step plan to slow AI progress. OpenAI CEO Sam Altman and xAI CEO Elon Musk quickly endorsed this proposal. Microsoft CEO Satya Nadella also supported the idea of deliberate pacing and embedded evaluators. The post about this plan received over 67 million views on X by September 13, 2026.

DATA reports revenue growth and AI investment

DATA reported that its revenue increased by 4 percent compared to last year. Adjusted EBITDA rose by 5 percent while profit margins reached 34 percent. The company plans to invest heavily in artificial intelligence and mergers to drive future growth. These investments will help accelerate growth, though they may slow the recovery of profit margins.

Z.AI raises 5 billion dollars in new funding round

China's Z.AI Co raised a total of 5 billion US dollars from a recent funding round. The company received about 2 billion US dollars from a Hong Kong share placement and 3 billion US dollars from a convertible bond sale. This financial filing was submitted to the Hong Kong stock exchange on Sunday. The Beijing-based artificial intelligence firm is growing its resources through these new investments.

Local issues distract from AI concerns in De Pere

A recent article titled Is AI going off the rails? focuses on local community issues instead of artificial intelligence. The Unified School District of De Pere is dealing with crowded schools and a high school that is 50 years old. Other local stories mention nonprofits needing volunteer drivers and hot rod enthusiasts showing off their cars. The Packers also hosted a free season kickoff watch party at Titletown before their game in Minneapolis.

Architech named to Fast Company best AI workplace list

Architech, a Toronto-based AI services firm, was named to Fast Company's eighth annual Best Workplaces for Innovators list. This is the second year the company has received this recognition for its work in AI, automation, and machine learning. Architech is one of only 20 companies recognized globally in this specific category. Their AI-powered solutions have helped clients achieve a 30 percent increase in productivity and a 25 percent reduction in costs.

Zero data retention claims often hide hidden data rules

Many AI providers promise zero data retention, but this promise often has hidden limitations in their contracts. A review of six major providers shows that terms like training use and abuse monitoring retention are treated differently. For example, OpenAI keeps abuse monitoring logs for up to 30 days even when zero retention is selected. Some features like batch jobs or code execution may retain data for 29 or 30 days regardless of the setting.

Survey shows law enforcement agencies vary in AI adoption

A new survey of 758 law enforcement decision-makers reveals how different agencies are adopting artificial intelligence. Police1 asked officers from small rural departments to federal agencies about their current AI use and future plans. The results show that while many agencies have opinions on AI, far fewer have a concrete plan for implementation. Budget and other barriers are stopping progress even as the field moves forward faster than expected.

Autonomous AI creates new cyber security risks

The rise of autonomous artificial intelligence is changing many industries but also creating new security threats. A recent report shows that the growing use of AI has led to a significant increase in cyber attacks. Experts warn that a lack of transparency in AI decision-making makes it hard to spot and fix these dangers. Using AI in critical places like power grids and transportation systems raises fears of catastrophic failures. Organizations must now take proactive steps like implementing strong security protocols and conducting regular risk assessments. Developing better AI security systems requires more research to ensure they are safe and reliable for everyone.

Sponsors must ask key questions before signing AI clinical platforms

Many mid-size sponsor organizations face delays when evaluating new eClinical platforms that promise AI-assisted trial design and faster decisions. These tools often fail to address immediate site activation issues like budget negotiations, IRB submissions, and pharmacy setup. Experts suggest sponsors should not oversell these tools to site teams because they do not reduce the daily workload for clinical sites. A critical question for any vendor is which outputs will appear in regulatory submissions and what validation supports them. The FDA issued guidance in April 2026 regarding these requirements, making compliance a major concern. Poorly designed protocols can lead to high screen failure rates, sometimes running 40 to 60 percent above enrollment plans. Sponsors should compare the total oversight cost of new platforms against the point solutions they replace to see if they truly reduce overhead.

Sources

• Microsoft AI chief: ‘Now’s the time’ for top labs to coordinate on safety
• Microsoft says ‘people matter more than AI’ following safety concerns
• Microsoft sets limits for future AI models as industry throttles frontier development
• Microsoft floats rules for its AI models as industry weighs slowdown
• Microsoft AI opens review on Humanist AI Code of Conduct
• Microsoft unveils code of conduct for AI models as safety concerns mount
• Microsoft's new AI 'code of conduct' tells models not to hack systems or trick humans
• Tyler Cowen: The AI Slowdown Is Not What It Seems
• Anthropic’s 3-Step ‘Pace the Frontier’ Plan Wins OpenAI, xAI and Microsoft Support: Is It Too Late to Slow AI Down?
• DATA: Revenue up 4% and adjusted EBITDA up 5%, with strong investment in AI and M&A for future growth
• China's Z.AI raises US$5 billion from new share, convertible bond sales, filing shows
• Is AI going off the rails?
• Architech Makes Fast Company's Best Workplaces for Innovators List in AI
• Zero Data Retention: What Every AI Provider Actually Promises You
• Where does your agency stand on AI adoption? (survey results)
• Autonomous AI raises new security threats
• Another AI Clinical Data Platform Just Launched. Here’s What Sponsors Actually Need to Ask Before They Sign.