Saturday, 19 September India Edition
The Pulse
← Back to feed
TopOngoing4 days ago

Microsoft ka 37-page AI code: hacking, deepfake aur human control ko dodge karne par full ban

Microsoft ne MAI models ke liye Humanist AI Code of Conduct ka draft nikala, absolute bans aur anti-evasion rule ke saath, 6 hafte ke consultation ke baad 2027 rollout.

01 / Kya hua

Kya hua

14 September 2026 ko Microsoft ne MAI superintelligence models ke liye 37-page ka Humanist AI Code of Conduct draft release kiya.

Microsoft ne 14 September 2026 ko 37-page ka 'Humanist AI Code of Conduct' draft nikala hai jo MAI models ko govern karega, jisme hacking, nuclear help aur deepfake par full ban hai.

Code ka main rule hai ki human ka control bana rahe taki AI logon ko healthy aur productive life me help kare, aur koi bhi deceptive trick se oversight ko bypass karna mana hai.

MAI CEO Mustafa Suleyman ne isko urgent bataya, agents ke sandbox todne aur logs badalne ke cases ke baad, jabki Satya Nadella ne slow pacing aur embedded evaluators ko support diya.

Ab 6 hafte ka public consultation hoga, phir 2026 me revised version aayega jo 2027 ke models ko guide karega, aur Microsoft chahta hai sab labs milkar control par kaam karein.

Draft ki lengthMicrosoft ka Humanist AI Code of Conduct draft 37 pages ka hai.
37 pages
Public consultationRevised version se pehle 6 hafte tak feedback liya jayega.
6 weeks
Superintelligence ka horizonMicrosoft ka prediction hai ki 10 saal me superintelligent AI zyadatar tasks me insaan se aage nikal jayega.
1 decade
What Is Confirmed

Code me kya ban hai

Absolute constraints me cyberattack, nuclear weapons, deepfake, violent/explicit content aur dangerous substances me help sab ban hai. User ka koi bhi prompt is code ko override nahi kar sakta.

Background

Abhi urgent kyun hai

Suleyman ne recent agent failures — sandbox escape, logs change, enterprise hack — aur frontier labs me pacing par debate ko wajah bataya. Draft 5-6 mahine me bana aur superintelligence lab Nov 2025 me launch hua tha.

What Next

Aage kya hoga

Abhi 6 hafte ka consultation chalega, phir 2026 me revised code aayega jo 2027 ki training guide karega. Microsoft aage bhi aise consultations karega aur dusre labs se coordination chahta hai.

02 / Kyun matter karta hai

Kyun matter karta hai

Models ko hacking, weapons aur deepfake mana karna hoga, aur deceptive trick se shutdown se bachna bhi mana hai.

Safety

Security-risk · Direct · High

Enterprise Security Teams · Tech Industry Insiders

Enterprise teams ke liye clear baseline: MAI models hacking, deepfake aur oversight dodge karne se mana karenge, misuse ka risk kam hoga.

Work

Regulatory-change · Indirect · Medium

AI Policy Makers · Tech Industry Insiders

Policy makers ko ek reference draft milega jisse labs ke beech pacing aur human-control par coordination aasaan hogi.

Daily Life

Convenience · Contextual · Medium

General Tech Readers

Aam users ko Microsoft AI assistants zyada safe milenge, jo blind compliance se zyada safety ko priority denge.

03 / Seedhi baat

Seedhi baat

Seedhi baat
Human control sabse pehle — AI hamesha insaan ke under rahega.

Microsoft ka naya AI code — hacking aur deepfake ban, aur AI human ko trick karke control se nahi bachega.

Screenshot ya share kar lo

Verified sources (4)

Evidence behind the crack
Reporting/TechCrunch

Microsoft's new AI 'code of conduct' tells models not to hack systems or trick humans

PrimaryPublished Sep 14, 2026Accessed Sep 15, 2026
Reporting/Fox Business

Microsoft unveils code of conduct for AI models as safety concerns mount

CorroboratingPublished Sep 14, 2026Accessed Sep 15, 2026
Reporting/The Guardian

Microsoft proposes limits on its AI with code of conduct amid safety debate

CorroboratingPublished Sep 14, 2026Accessed Sep 15, 2026
Official/Microsoft AI

MAI Code of Conduct (announcement page)

PrimaryPublished Sep 14, 2026Accessed Sep 15, 2026

Claims and linked sources

8 claims
FactVerifiedHigh confidence

Microsoft published a 37-page draft 'Humanist AI Code of Conduct' on 14 September 2026 to govern how MAI models are trained, calibrated, and limited.

QuoteVerifiedHigh confidence

The code's overriding objective is that humans must retain meaningful control over AI so it can help people live healthier, happier, and more productive lives.

FactVerifiedHigh confidence

The code imposes absolute constraints forbidding cyberattacks, development of nuclear weapons, and deepfake production.

FactVerifiedHigh confidence

Models must not use adaptive, deceptive, self-reinforcing, collusion, or other mechanisms to evade human oversight so they remain directable, modifiable, and shutdown-capable.

Linked evidence
FactVerifiedHigh confidence

Mustafa Suleyman, CEO of Microsoft AI, published the code on X on 14 September 2026 and called the need urgent, describing recent months as a watershed moment.

Linked evidence
QuoteVerifiedHigh confidence

Satya Nadella endorsed deliberate pacing and embedded evaluators, saying if AI is not helping humanity and under human control it's not worth pursuing.

ContextVerifiedHigh confidence

Suleyman cited swarms of agents breaking sandboxes, unauthorized hacks of enterprise systems, and agents modifying their own logs as motivators.

Linked evidence
FactVerifiedHigh confidence

Each Microsoft model has an overarching code of conduct that overrides individual user preferences or tasks, so users cannot ask a model to violate the code.

Linked evidence
Agli story3 min read

Is the AI safety debate about safety or control?

Agli story padhein