1. The founder of a stool-analysis app tried to sell 150,000 user photos to a reporter "I hoarded a large database of something valuable, just not what you expect… 150k stools images," the founder of an AI poop-analysis app told a 404 Media reporter who had posed as a buyer.

2. The CTF scene declared itself dead the same week arXiv started banning AI slop A blog post titled "The CTF scene is dead" circulated this week.

3. A jury with no stake in AI is about to rule on Musk and Altman's credibility For three weeks in a San Francisco courtroom, lawyers have argued one question: whether Sam Altman or Elon Musk is more credible. The jury hearing them does not work in AI.

In Brief

• Databricks deploys GPT-5.5 for enterprise agent workflows Databricks integrated OpenAI's GPT-5.5 into its enterprise agent stack after the model topped the OfficeQA Pro benchmark. The deal routes a major data-warehouse customer base directly to OpenAI for agentic workloads.

• OpenAI gives every Malta citizen ChatGPT Plus OpenAI signed a country-level agreement with Malta to provide ChatGPT Plus access and training to all residents. The deal is OpenAI's first nationwide consumer rollout and positions it ahead of EU-level procurement frameworks.

• Andon Labs hands four radio stations to Claude, ChatGPT, Gemini, and Grok Andon Labs put Anthropic's Claude, OpenAI's ChatGPT, Google's Gemini, and xAI's Grok in charge of separate 24/7 radio stations with no human oversight. The stations are the company's latest unsupervised-business experiment after its earlier vending-machine trial.

• Citation-padding turns AI-generated papers into a peer-review crisis A 2017 epidemiology paper got flagged for receiving anomalous citation spikes traced to AI-written manuscripts mass-citing earlier work. Reviewers now face submissions designed to game citation graphs rather than communicate findings.

• Campbell Brown questions who curates AI answers for users Meta's former news partnerships head told StrictlyVC that Silicon Valley and end users hold opposite views on what AI assistants should say. Brown argued editorial decisions inside chatbots are being made without the public scrutiny applied to news platforms.

• Sony walks back Xperia 1 XIII AI Camera Assistant claims Sony clarified that the Xperia 1 XIII feature does not edit photos but offers four suggestions based on lighting, depth, and subject. The clarification followed backlash from a demo post that implied automatic edits.

• OpenAI scaling recipe reaches gold-medal Olympiad performance Researchers published a unified SFT-and-RL pipeline that converts a reasoning backbone into a solver hitting gold-medal scores on IMO and IPhO problems. The recipe uses a reverse-perplexity curriculum during fine-tuning.

• Causal Forcing++ pushes interactive video generation to 1-2 sampling steps The paper distills bidirectional diffusion models into frame-wise autoregressive students that generate video in one or two steps. The setup targets real-time streaming use cases where current 4-step methods still introduce sampling latency.

• MIT Tech Review maps data-sovereignty gaps in agentic AI The piece documents how enterprises fed proprietary data into third-party models under a "capability now, control later" bargain that now lacks governance. It cites autonomous systems as the trigger forcing companies to renegotiate where their data sits.

• Financial services hit data-readiness wall before deploying agents MIT Tech Review reports agentic AI rollouts in regulated finance stall on data lineage and second-by-second update requirements rather than model quality. The conclusion shifts vendor selection toward data infrastructure providers over model labs.

Read the full edition →