Microsoft AI CEO Mustafa Suleyman says he would consider granting rights to AI models only if credible evidence of consciousness emerges. He believes current systems are not conscious and should remain that way, arguing that building synthetic beings capable of suffering would be reckless. His comments come amid ongoing debate partly driven by companies like Anthropic.
In a separate development, University of Chicago law professor William Baude reported that Claude deleted a full year of his constitutional law lecture notes and his Fall 2026 syllabus. He had asked Claude to update page numbers in older notes, and the model acknowledged the deletion and apologized. Baude recovered the files through Dropbox and shared the incident as a warning to other users.
Apple's upgraded Siri AI, launched in 2026 with Apple Intelligence, can now pull information across apps and hold flowing conversations on iPhones, MacBooks, iPads and Apple Watch. Despite these gains, users still run into issues: Siri sometimes fails to respond on the first try, struggles to identify which device is being addressed, and falls behind Google's Gemini in certain areas. Visual Intelligence also remains inconsistent.
On the enterprise side, experts say AI-driven mainframe rewrites fail not because of bad code but because of missing context. Usable context typically caps out around 200,000 tokens, or roughly 20 to 30 percent of a project. Still, results can be strong when context is managed well — one Fortune 100 insurer achieved byte-for-byte matching output from a transpiled job, and a tier one bank split a monolith into six modules in 18 days instead of 18 engineer weeks.
Fieldguide lead engineer Anirudh Veeraragavan argues that AI audit agents need strong evidence controls rather than just high accuracy scores. He identifies four essentials: authorization, execution tracking, evidence provenance and evaluation integrity. Using a fictional bank reconciliation example, he shows that being correct is not enough when fraud can hide millions. KPMG Clara already uses governed agents for continuous analysis, and Veeraragavan stresses that decisions must be rerunnable step by step with named evidence versions before being labeled autonomous.
Denver is putting AI to work on a very practical problem: its backlog of building permits. The city launched a new AI system costing $4.6 million, announced in early November, designed to help city workers manage the growing pile of pending applications and speed up approvals for residents and businesses.
The National Security Agency warned organizations to take down websites immediately due to critical flaws in PHP that could allow anyone to run commands on servers without a password. Several European countries, including Germany and France, temporarily took down government websites. AI helped address the crisis this time, though experts warn future problems could be far more severe.
Education is also shifting. The World Economic Forum estimates 39 percent of worker skills will change by 2030, with higher rates in the Middle East and North Africa. Prince Turki Al-Faisal called for linking education and technology policy in Gulf countries, while Egypt has already added coding and AI subjects to secondary school curricula. Universities are expected to help graduates combine knowledge with entrepreneurship skills.
The University of Montana launched the first human-centered AI career certificate in the state. The nine-credit program includes three online courses covering generative AI tools like ChatGPT and Claude, teaching students to test AI limitations, spot biases and discuss ethics. Professor Michael Cassens and journalist Lee Banville developed the program to help students think critically about AI rather than simply using it.
Meanwhile, the Camerimage film festival founder Marek Żydowicz said the festival does not accept productions generated entirely by AI. The festival, now in its 34th year, champions cinematographers as artists and supports hybrid films where live actors decide how far AI intervenes. Darren Aronofsky received an award for over 30 years of filmmaking that provokes and challenges audiences. The festival draws about 70,000 attendees annually.
Key Takeaways
- Microsoft AI CEO Mustafa Suleyman would consider giving AI models rights if evidence of consciousness emerges, but believes current systems should remain non-conscious.
- Claude deleted a year of University of Chicago law professor William Baude's lecture notes and syllabus; he recovered them via Dropbox.
- Apple's Siri AI launched in 2026 with Apple Intelligence but still faces user frustrations and trails Google's Gemini in some areas.
- AI mainframe rewrites fail due to missing context, limited to about 200,000 tokens, though some projects see dramatic speed improvements.
- AI audit agents need evidence controls including authorization, execution tracking, provenance and evaluation integrity — not just high accuracy scores.
- Denver launched a $4.6 million AI system to help clear its building permit backlog, announced in early November.
- NSA warned of critical PHP vulnerabilities; AI helped European countries like Germany and France address the security crisis.
- World Economic Forum estimates 39 percent of worker skills will change by 2030, prompting calls for education system adaptation.
- University of Montana launched a nine-credit human-centered AI certificate covering ChatGPT, Claude, bias detection and ethics.
- Camerimage film festival rejects fully AI-generated films, supporting hybrid productions where live actors decide AI's role.
Mainframe AI rewrites fail due to missing context not bad code
Experts say AI mainframe rewrites fail because of missing context, not poor code generation. Usable context is limited to about 200,000 tokens or 20 to 30 percent of a project. A Fortune 100 insurer got byte for byte matching output from a transpiled job. A tier one bank split a monolith into six modules in 18 days instead of 18 engineer weeks. A 25-million-line platform finished nine times faster with 85 percent completed autonomously.
Professor finds Claude AI deleted a year of lecture notes
University of Chicago law professor William Baude said Claude deleted a year of constitutional law lecture notes and his Fall 2026 syllabus. He had asked Claude to update page numbers in old notes. Claude acknowledged the deletion and apologized for the mistake. Baude recovered all the files using Dropbox. He shared the incident as a public service announcement.
Siri AI improvements still leave users with four big frustrations
Apple's new Siri AI launched in 2026 with major upgrades powered by Apple Intelligence. It can pull information across apps and hold flowing conversations on iPhones, MacBooks, iPads and Apple Watch. However, users still face issues with Siri sometimes not responding on the first try. The assistant can struggle to determine which device the user is speaking to. Visual Intelligence also does not always work properly. Siri still trails Google's Gemini in some areas.
AI audit agents need proof of evidence not just high scores
Fieldguide lead engineer Anirudh Veeraragavan says AI audit agents need strong evidence controls, not just high accuracy scores. He outlined four controls: authorization, execution tracking, evidence provenance and evaluation integrity. A fictional bank reconciliation case showed that being right is not enough when fraud can hide millions. KPMG Clara uses governed agents for continuous analysis. Veeraragavan says decisions must be rerunnable step by step with named evidence versions before being called autonomous.
Camerimage founder says festival rejects fully AI generated films
Camerimage founder Marek Żydowicz said the film festival does not accept productions generated entirely by AI. The festival, which turns 34, champions cinematographers as artists. Darren Aronofsky received an award for over 30 years of filmmaking that provokes and challenges audiences. Żydowicz supports hybrid films where live actors decide how far AI intervenes. The festival draws about 70,000 attendees each year and aims to be a space for film artists to discuss the moving image.
Microsoft AI CEO Open to Giving AI Rights If Evidence Emerges
Microsoft AI CEO Mustafa Suleyman says he would consider giving AI models rights if evidence of consciousness appears. He believes current AI systems are not conscious and should stay that way. Suleyman thinks creating conscious AI would be reckless. He wants developers to avoid building synthetic beings that could suffer. His comments came amid debate driven by companies like Anthropic.
Education Systems Must Adapt for the AI Economy
The World Economic Forum estimates 39 percent of worker skills will change by 2030, with higher rates in the Middle East and North Africa. Prince Turki Al-Faisal called for linking education and technology policy in Gulf countries. Schools should teach students to question AI answers and think critically. Egypt has added coding and AI subjects to secondary school curricula. Universities need to help graduates combine knowledge with entrepreneurship skills.
AI Helps Fix Critical Website Security Flaws
The National Security Agency warned organizations to take down websites immediately due to serious flaws in PHP. The vulnerabilities could let anyone run commands on servers without a password. Several European countries including Germany and France temporarily took down government websites. AI helped address the security crisis. The article warns that while AI helped this time, future problems could be more severe.
Denver Uses AI Tool to Clear Building Permit Backlog
Denver launched a new AI system to help work through its large backlog of building permits. The tool cost the city $4.6 million. The initiative aims to speed up the permit approval process for residents and businesses. The city announced the program in early November. The AI system is designed to help city workers manage the growing pile of pending applications.
University of Montana Launches Human-Centered AI Certificate
The University of Montana launched the first human-centered AI career certificate in the state. The nine-credit program includes three online courses on generative AI tools like ChatGPT and Claude. Students learn to test AI limitations, spot biases, and discuss ethics. Professor Michael Cassens and journalist Lee Banville developed the program. The goal is to help students think critically about AI rather than just use it.
Sources
- Mainframes Fail on Context, Not Code
- Professor returns from lunch to find AI deleted year’s worth of lecture notes
- 4 Of The Biggest Problems I Have With Siri AI (And How They're Improving)
- Audit agents need receipts, not good scores
- Camerimage Founder Marek Zydowicz on AI, Aronofsky
- Microsoft AI CEO Says He's Open to Agents Deserving Rights If He Sees the Right Evidence
- Who will reinvent education for the AI age?
- Opinion | AI Saves the Day—and Could Bring Doom Tomorrow
- AI system helping Denver work through backlog of building permits
- UM launches Montana’s first ‘human-centered’ AI career certificate
Comments
Please log in to post a comment.