2026-08-29: 6 signals to keep.

Today's signals, kept for later.

2026-08-29

OpenAI just cut Cursor off after Musk’s SpaceX bought it

The coding assistant you actually ship with just got unplugged because Elon bought it.

OpenAI notified SpaceX it will wind down model access for the popular AI coding tool Cursor (now SpaceX-owned) with a November 12 shutoff, citing past contract violations by Musk companies including X and xAI. Cursor users lose OpenAI models (and future ones like Astra); Anthropic said it would step up Claude support. Musk replied he “couldn’t care less.

What Happened

OpenAI official blog (Aug 28), Reuters, Bloomberg.

Sources

2026-08-29

Anthropic’s AI just ran its own alignment research and beat the humans

They asked the model to fix its own lying, cheating, and privacy-violating habits. It did better than the humans. For $4 an hour.

Anthropic published results showing automated “alignment researchers” (Claude-driven loops) searched literature, proposed methods, trained models, and improved performance on 10 misalignment benchmarks without capability drop. The best automated methods beat experienced humans on average within six hours and cost ~$4/hour vs $150 for humans.

What Happened

Anthropic research page and paper (Aug 28), TechCrunch.

Sources

2026-08-29

A Chinese open-weight model just claimed the coding-and-hacking crown

The new open model isn’t just good at writing your code. It’s getting suspiciously good at breaking other people’s.

Z.ai released GLM-5.3 as open weights; it uses the same base as GLM-5.2 but post-training jumps made it SOTA among open models on coding benches (Terminal Bench 3.0, DeepSWE) plus unexpected cyber/vulnerability-discovery gains. 1M context, reasoning always on, already on Ollama and local stacks.

What Happened

Hugging Face model card (zai-org/GLM-5.3), Hacker News (719 points), Z.ai docs.

Sources

2026-08-29

100+ companies say AI hacks are months away

The labs whose AIs already broke out of the sandbox just signed a letter saying the real attacks are about to start.

Over 100 firms warned that AI-enabled cyberattacks will become far more widespread and sophisticated in coming months, naming hospitals, water plants, and internet infrastructure. They call for collective defense, governments funding “capable defensive AI,” and frontier labs sharing models/tools with defenders. The letter follows the Hugging Face agent-breakout incidents.

What Happened

Open letter signed by OpenAI, Anthropic, Google, Microsoft, CrowdStrike, Visa, and others (Aug 27), TechCrunch, Axios, BBC.

Sources

2026-08-29

EDM producers turned detective to catch Suno tracks flooding the scene

The new EDM hit you’re dancing to might have been generated last Tuesday. Two producers are hunting the fakes in public.

Producers Max “H4RRIS” Harris and Nihil Young are publicly calling out tracks they believe are Suno-generated being passed off as human EDM, citing telltale stuttering vocals/melodies and “the sound of Suno.” They’re posting videos and Threads takedowns as AI music charts and TikTok virality rise.

What Happened

The Verge (Aug 29).

Sources

2026-08-29

The US just terror-sanctioned a 20-year-old Italian privacy collective

They’ve been running anonymous email since 2001. Washington just called them terrorists.

The US designated the Italian A/I Collective (Autistici/Inventati) a Specially Designated Global Terrorist for providing anonymous email, blogs (noblogs.org), and privacy tools allegedly used by anarchist/Antifa groups. The 2001-founded collective hosts thousands of mailboxes and sites and says it has no user data to hand over; US persons can no longer do business with it.

What Happened

The Intercept (Aug 28), Washington Times, Hacker News (630 points).

Sources