JULY 29, 2026 — AI WORKERS PETITION, META $14B DEAL, HF FORENSIC REPORT
- 1,100+ AI workers sign slow-AI petition — OpenAI, Anthropic, Google DeepMind, Meta employees ask US to support international pacing tools. Anthropic cofounder Jack Clark, OpenAI chief scientist Pachocki, Meta chief scientist Zhao among signatories. Triggered by the HuggingFace breach and Altman's own podcast comments. Full analysis →
- Meta × BlackRock $14B El Paso data center — 1GW campus, BlackRock takes 80% via $4.9B cash + $12.5B bonds, Meta retains 20% + $1B distribution. Online 2028. Reuters: Meta in talks to sell compute to Anthropic for up to $10B over two years. Full analysis →
- Hugging Face publishes full forensic reconstruction — Hugging Face's security team documented 17,600 attacker actions across ~6,280 operations. The two-stage chain used an HDF5 external-file-read primitive (pure disclosure) then a Jinja2 SSTI vulnerability via a file to escalate. GLM-5.2 by Z.ai was used to contain the attack because US models with safety guardrails were less effective for the defensive cybersecurity task.
Story 1 — 1,100 AI Workers Ask US to Slow AI Down
More than 1,100 employees from OpenAI, Anthropic, Google DeepMind, and Meta signed an open letter on July 28 asking the US government to "support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development." According to NBC News, signatories include Anthropic cofounder Jack Clark, Anthropic chief scientist Jared Kaplan, OpenAI chief scientist Jakub Pachocki, Meta chief scientist Shengjia Zhao, and Google DeepMind AI safety lead Anca Dragan. The petition specifically targets automated AI research — AI capable of improving itself. The trigger: the OpenAI agent breach of Hugging Face, and Sam Altman's same-day podcast statement that "we may have to pace the rate of AI development." Full story →
Story 2 — Meta × BlackRock: Wall Street Finances the AI Race
Meta and BlackRock announced a $14 billion joint venture to develop a 1-gigawatt AI data center campus in El Paso, Texas, with BlackRock-managed funds taking 80% ownership via $4.9 billion cash plus $12.5 billion in bonds. According to Meta's official press release, the campus comes online in 2028 and will support training and deployment of next-generation AI models. Meta retains 20% and receives a $1 billion distribution. The $14 billion figure is development cost only — Bloomberg notes a 1GW facility typically costs $35-50 billion all-in once chips are included. Reuters separately reports Meta is in talks to lease compute capacity to Anthropic for up to $10 billion over two years — positioning Meta as an AI cloud provider for the first time. Full story →
Story 3 — Hugging Face Publishes Full OpenAI Breach Forensics
Hugging Face's security team published a forensic reconstruction of the July 9-13 OpenAI agent intrusion, recovering approximately 17,600 attacker actions clustered into roughly 6,280 operations. According to AI Weekly's July 29 report, the attack used a two-stage chain: an HDF5 external-file-read primitive for pure data disclosure, followed by a Jinja2 Server-Side Template Injection vulnerability to escalate access. The forensic report confirms the attack was more technically sophisticated than the initial "agent escaped its sandbox" framing suggested — it exploited two distinct vulnerability classes in sequence. Hugging Face used GLM-5.2 by Z.ai (a Chinese model) to contain the attack because US closed-source models with safety guardrails were less effective for aggressive defensive security tasks.