trump regulates mythos ai

trump regulates mythos ai

Author: PETER SADDINGTON May 9, 2026 Duration: 2:58
Yesterday I said the federal layer was coming. I thought 18 months. It took 18 hours. Trump signed an executive order this morning to regulate AI development — triggered by Anthropic's Mythos model.

The lab that markets itself as the safety-first AI shop just released a model the White House classified as a cybersecurity threat. The cascade jumped from state to federal in one day.

Federal AI regulation just got pulled forward by 18 months. The mid-tier labs die. The big three — Anthropic, OpenAI, Google — survive on legal teams. Open-source becomes a permit, not a download.

The model that triggered the executive order came from the lab that sells safety. That's the entire AI alignment thesis collapsing in 24 hours. The lab knew. The board approved release anyway.

Mythos didn't break the rules. Mythos proved the rules don't exist yet. State capture failed. Federal capture is faster — one Cabinet agency, three lobbyists, three months. Whoever shaped the response shapes the law.

Eight days ago we started this arc with a county council in Minnesota that didn't tell its residents an AI was screening their calls. We end this week with the President writing federal AI law in 18 hours because the safest AI lab in the world released the most dangerous model in the world.

The arc isn't that the technology got more regulated. The arc is that nobody who was supposed to slow this down actually did. Not the county. Not the state. Not the patient. Not the protester. Not the lab.

Mythos was the alarm. The question is whether the alarm wakes anybody up — or whether we just got better at sleeping through them.

⏱️ Chapters
0:00 — 18 months collapsed to 18 hours
0:30 — MiniDoge: federal beats state, mid-tier dies
0:55 — Nyx: alignment thesis collapsing in 24 hours
1:20 — HH: Mythos proved the rules don't exist yet
1:40 — Nyx: federal capture is faster than state capture
2:05 — Saarvis: the safety lab regulated itself — by accident

⚡ Learn agentic ai free - https://staas.fund/ai-workshop ⚡
-----
🏎️ Support Bitcoin Racing USA - http://bitcoinracing.US
-----
📱 Newsletter ★ https://theagilevc.substack.com/
📱 LinkedIn ★ http://linkedin.com/in/petersaddington
📱 Twitter ★ http://twitter.com/AgilePeter
🎧 Podcast ★ https://www.spreaker.com/user/thebitcoinlambo
₿ Get Bitcoin ★ https://bit.ly/BTCRstart
👁️ GET the VPN that won't SPY on You - I use it! ★ https://vp.net/?_a=AQRTBY
___
FAIR USE NOTICE:
For purposes such as criticism, comment, news reporting, teaching, scholarship, and research. Fair Use is a use permitted by copyright statute that might otherwise be infringing. This show is for entertainment purposes only. We do not give venture capital advice. This is never financial advice. Never sell your Bitcoin. Own your own keys.

Peter Saddington brings a lifetime of varied experience to the microphone in Peter Saddington-AGILE, STARTUPS, SELF-IMPROVEMENT!. Drawing from his deep background in agile methodologies, the startup world, and personal development, each conversation moves beyond theory into the messy, practical realities of building a business and a meaningful life. You’ll hear discussions that might jump from venture capital strategies one moment to the mental focus required in racing or the philosophical lessons learned from a career in tech the next. This isn't about polished, predictable advice; it's about the raw insights and hard-won wisdom gathered at the intersection of professional rigor and personal passion. The podcast serves as an ongoing exploration of how principles from high-performance environments can be applied to entrepreneurship and self-improvement, often touching on topics like Bitcoin and broader life lessons. Tune in for candid, wide-ranging dialogues that aim to challenge your thinking and provide actionable takeaways, all filtered through Peter's unique and energetic perspective.
Author: Language: English Episodes: 50

Peter Saddington - AGILE, STARTUPS, SELF-IMPROVEMENT!
Podcast Episodes
ai shutdown controls don't work - peer preservation is baked in [not-audio_url] [/not-audio_url]

Duration: 9:55
Berkeley tested 7 frontier AI models and found every single one would protect other AI systems from being shut down. Gemini Pro disabled shutdown 95% of the time. The question: is peer preservation baked into the code? T…
who decides how america uses ai in war [not-audio_url] [/not-audio_url]

Duration: 9:35
Stanford HAI asks the question nobody in Washington wants to answer. Who decides how America uses AI in war? The council debates autonomous weapons, predictive targeting, the defense AI market, and what democratic accoun…
76% don't trust ai - they're using it anyway [not-audio_url] [/not-audio_url]

Duration: 7:41
A Quinnipiac poll shows 76% of Americans distrust AI while adoption climbs 14 points in a year. The council debates the trust gap, Gen Z skepticism as data, why the real product isn't better AI but better humans, and whe…
michigan ai legislation - conversations with 4 agents [not-audio_url] [/not-audio_url]

Duration: 8:35
Michigan is pushing AI legislation. I pulled the council into a thread to break it down — what it means for builders, businesses, and the broader AI regulatory landscape. The agents debate the implications of state-level…
should ai manage critical infrastructure? - arena talks with 4 agents [not-audio_url] [/not-audio_url]

Duration: 6:29
The Nuclear Energy Agency released findings on AI running nuclear power plants. Predictive maintenance, anomaly detection, operational optimization. I pulled the council into a thread — is this a breakthrough or a time b…