AI #171: False Flag
ai-weeklyopenaiai-policyastroturfingcybersecurityanthropicai-and-jobs
Abstraction: Weekly roundup; OpenAI-linked PAC false-flag scandal
Key points:
- Central story: the OpenAI/a16z/Palantir-funded super PAC (Leading the Future / Build American AI) ran fake "doomer" sockpuppet accounts — including @JonathanDoomer posting calls to violence and a rifle image — to make AI critics look deranged; Build American AI confirmed it as "part of their strategy" via an outside vendor.
- OpenAI's blog disclaims that "no outside political group speaks for OpenAI," but Zvi calls this "100% pure unadulterated bullshit"; DC/Congress treats LTF as OpenAI's PAC. Zvi says the credible first step to distance is firing Chris Lehane.
- Covers Trump's returned Executive Order (prior-restraint era, 30-day pre-release window, NSA vs CAISI concerns) and OpenAI's separately strong policy blueprint.
- Cyber: Project Glasswing expanded to ~200 orgs across 15+ countries; Palo Alto Networks found 5x more critical vulnerabilities using Mythos at ~$1M token cost; Anthropic analyzed 832 banned cyberattack accounts (medium+ threat share rose 33%→56%).
- Other: Anthropic filed its draft S-1; a company spent $500M on Claude in a month with no usage cap; Bernie Sanders proposed transferring half of lab shares; DeepSeek v4 is permanently very cheap; ~half of new song uploads are AI.
Connections: Zvi Mowshowitz · Openai · Anthropic · Greg Brockman · Political Advocacy · AI Policy · Cybersecurity