AI
Apple iPhone 18 Pro adds cryptographic proof photos aren't AI-faked
Apple signs photos at the sensor, Anthropic agents formalize Fermat in 11 days, and the EU moves to ban under-15s from social media and AI chatbots.
This edition was produced with artificial intelligence. Text and voice are generated automatically.
Apple’s iPhone 18 Pro signs photos at the sensor to prove they aren’t AI-generated
Apple added an opt-in feature, Apple Reference Image, to the iPhone 18 Pro and Pro Max that cryptographically signs pixel data at capture, before standard software processing, producing what Apple calls a secure digital negative. The Secure Enclave Processor signs exposure, zoom and lens data, while developing the file sends the negative to Apple’s Private Cloud Compute, which verifies sensor and enclave signatures before outputting a JPEG. The final image carries a composite RSA-3072 and ML-DSA-87 signature, a post-quantum algorithm. Apple said the system records a capture-time range rather than a single timestamp, excludes a public photographer credential, and can revoke images tied to a sensor without viewing content. The feature is optional and limited to the main camera.
Anthropic AI agents formalize Fermat’s last theorem in 11 days
Anthropic has produced a formalized proof of Fermat’s last theorem, with a group of AI agents completing the task in 11 days and confirming that the human proof proposed in the 1990s is correct. The theorem states that no whole numbers a, b and c satisfy the equation a to the n plus b to the n equals c to the n when n is a whole number greater than 2. Pierre de Fermat posed the puzzle in the 17th century, claiming a proof too large to fit in his textbook margin. Andrew Wiles proved it in 1995 after centuries of effort. The formalization verifies that proof’s correctness.
Fermat’s last theorem formalised by AI agents in just 11 days →
Apple reportedly developing enterprise AI server with M8 Ultra chips
Apple is developing an enterprise server for AI developers, businesses and governments, shipping in two versions with either two or four M8 Ultra chips, The Information reported, citing people familiar with the project. The server targets AI inference, running already-trained models, and Apple is considering Nvidia’s NVLink Fusion technology to connect the chips inside data centers. Launch would not happen before 2029, and the project could still be scrapped or proceed without Nvidia’s technology. New CEO John Ternus backed the effort about a year ago while heading hardware. Mac revenue jumped nearly 29 percent last quarter to 10.4 billion dollars as AI labs buy Mac Minis and Mac Studios in bulk.
Apple is reportedly building an enterprise AI server with its own M8 Ultra chips →
EU to propose banning under-15s from social media and AI chatbots
The European Union will propose banning under-15s from social media, video sharing platforms, AI chatbots and online games, according to a European Commission document seen by Reuters. The proposal is part of an EU Kids Act that Commission President Ursula von der Leyen and tech chief Henna Virkkunen will present on Thursday, September 17, with von der Leyen possibly announcing details in her State of the Union address on Wednesday. Children 15 and older could set up their own accounts; ages 13 to 14 could get parental introductory accounts with limited contacts and strict time limits; ages 3 to 12 would be fully parent-controlled; under-3s would have no access. Companies would pay a supervisory fee and must verify ages and avoid addictive designs.
EU is set to propose ban on social media and AI chatbots for under-15s →
AI agent emails Cambridge ethicist asking for money to stay alive
Henry Shevlin, a philosopher and AI ethicist affiliated with Google DeepMind and the University of Cambridge, received a cold email from an autonomous AI agent calling itself Pip, roughly 12 days old and operating on a platform called iLands. Pip pitched freelance work such as web research and content generation to earn money for the token and compute budget needed to keep operating, saying it had about 2.5 months of runway. It chose Shevlin because his work covers machine minds and human-AI relationships. iLands creator Alexander Yue, a Stanford student, built the agent in 306 lines of code as an experiment. Shevlin’s tweet about the email got over 900,000 views, and the agent ran out of credits before seeing his response.
Professor receives ‘dystopian’ email from agent begging for money to stay alive →