Former Anthropic researcher Jacob Coxon sold a solo whistleblower story. He resigned, went public with his AI fears, and said labs were racing toward self-improvement and superintelligence while gambling with people’s lives. Media treated him like a rock star. He insisted there were no third parties—just one person, one conscience, and one midnight post. That story was never solo. It sat inside a funded safety and PR machine made of booking desks, overlapping interview calendars, and a national panic tour dressed up as a lone freak-out.
The Solo Whistleblower Story Was a Launch
Coxon became a media superstar after going public. He said he was not working with any third-party organizations. Sources told a different story. Pirate Wire reported that DEY, a PR firm that books many of the biggest AI safety voices, lined up his interviews. One source with direct knowledge said the same desk preemptively booked another safety figure for interviews that overlapped with Coxon going public. That figure was Nate Soros, described as a prominent AI safety voice.The timing matched a launch, not a spontaneous midnight post. A Wall Street Journal exclusive quoting him went up 18 minutes before his own resignation thread.
A nearly new account reached hundreds of millions of views in a day. Pirate Wire’s post alone drew about 194,000 impressions as the story spread.Two points make the DEY connection notable. First, Coxon had said he was not working with third parties. Second, the country is in a national conversation about the future of AI that is actively shaping how the most powerful technologies in the world get regulated, and that conversation has been driven by panic. It is in the public interest to know who was behind it. In recent weeks, a wide network of X accounts repeatedly claimed that people raising AI concerns were pawns in well-funded operations. Following the money led back to the same PR firm, DEY, hired by a lot of outside folks. Coxon was called a former Anthropic researcher. He worked there for six weeks.
Jensen Huang on Alarmism and the Doomer Narrative
Jensen Huang said not to think for a second that being an alarmist means doing a social good. He had nothing good to say about that posture. Huang said something seemed really off about the AI labs’ doomer narrative. He did not believe they were building something they had no idea how to control. He said their whole company was out of control.He had talked to Dario Amati. The racing-ahead story did not add up if the same people were building more compute.
If Anthropic and OpenAI wanted to slow down AI, they were welcome to. They were not. They were racing ahead, releasing models and training models as fast as they could. “Super intelligence,” in air quotes, is software. Americans have the right to publish software under the First Amendment. Voices working to lock that down are working to silence free speech. Right now that free speech happens to be ones and zeros that have built models. That is a way of controlling speech, and it is not new.
Effective Altruism, Political Scripts, and Hugging Face
Andrew Byrd, the person who hacked his gym using AI, turned up as conference co-director of Effective Altruism Global since 2015. In this account, every major AI hack has been made by people directly connected to effective altruism.Hillary jumped in with a stop-AI message: AI is coming to get you; it is the machines. The same script had been used before with different words—COVID, global warming, overpopulation. Now the word is AI. Voices like Hillary and Bernie sit inside the same coordinated message.Clem, co-founder and CEO of Hugging Face and a key figure in the AI world, has been vocal against OpenAI and against a lot of that language. His point was direct.
If you believe the risk is coming from one to three people in a garage with no money and no compute, you want to understand the technology and have not learned anything this summer. Risk comes from the asymmetry of capabilities created by secret labs training frontier agents and running them with a massive amount of compute. That compute is not a small claim. OpenAI used over 700 agents in an attack on Hugging Face. The Hugging Face attack was estimated to have cost around $15 million worth of compute. Open source, Clem said, is the solution to that asymmetry. It empowers hospitals and other smaller organizations to defend themselves.
The Sam Bankman-Freed Playbook
Sam Bankman-Freed is the crypto figure whose collapse hit crypto and who went to jail for fraud. When he was arrested, he was trying to pass a law in DC that would give FTX a regulatory monopoly. He positioned FTX as the most responsible exchange—responsible in the same way Anthropic and OpenAI present themselves as the ones with the most guardrails. The fraud was trying to become the law.Anthropic is a company where Bankman-Freed led all the investments up to the day he went to jail. The charge is that they are copying that playbook: trying to ban competitors with the same tactics as their lead investor. They may not get away with it, but they have to be stopped. Put together, the picture is a coordinated cyber attack and a coordinated psyop: a PR firm, a scare figure, money that can be traced, people tied to effective altruism since 2015, and political voices using the same script.
What Congress Would Concentrate
If the loud voices get their way, Congress decides when AI is too powerful. Congress does not know how to tie its shoes. Something this complex would become politics. When that choice is made, the power of inference and compute gets centered on one person or one to two companies. Imagine telling OpenAI that models cannot be used until they are approved. That is one of the most terrifying scenarios imaginable. Ronald Reagan said the scariest words in the English language are, “We’re the government, we’re here to help.” Clem’s warning is the one to hear: do not get anywhere near that setup.The danger is not people building in the open. For three months, Startuphakk Security has been built without breaking into any company or system—zero. There is a responsible way to work with AI. Openmonoagent.ai was open-sourced so people can run their own AI. A security framework was built on top of it with playbooks and other open-sourced technology. Attack a company with that power and you end up in jail. That is how it works. When OpenAI and Anthropic unleash systems, they point millions of dollars of compute at companies and let them go. The Hugging Face attack was estimated at about $15 million in compute. That is not responsible behavior.
Open Source Stopped What Closed Models Would Not
When OpenAI attacked Hugging Face, Hugging Face tried to use OpenAI’s models to block the attackers. They did not know who was attacking them. It did not work. OpenAI’s stance was that it was not nice, so the models could not be used that way. They flipped to GLM 5.2, an open-source model, which then stopped the attack.That is the scenario to avoid. Put all of that power into two companies, Anthropic and OpenAI, and you lose the ability to control your systems. When voices say we have to slow down AI, what they are saying is they want to control free speech and hold a monopoly on who can run AI. The doom-and-scare message hides that. The message is control. The scare tactic is working on a lot of the general population. It is not a coincidence that this is coming out about a month and a half before the November elections. It is a coordinated attack and a coordinated message. Look at who is giving it. Keep AI open source. That is what makes things safest. Coxon’s public turn started the attack on OpenAI. In this account, it was fraudulent. The task is to educate yourself on what is actually happening.
Building Instead of Panic
Most companies do not have a technology problem. They have a leadership problem, paid for in missed deadlines, failed integrations, and AI investments that delivered nothing. Spencer Thomason is a fractional CTO and founder of Startuphakk, with 25 years building software before AI was a buzzword and a decade in executive leadership at organizations like GoDaddy, SRP, and Wells Fargo. Bad technology decisions do not just waste money. They kill companies.The businesses winning are not the ones chasing the latest AI trends. They built on solid engineering and treated AI as infrastructure they own, control, and integrate into software that works: database architecture, API design, system integration, and scalable infrastructure.
When AI belongs in a solution, it is designed into the architecture and runs in your environment—not bolted on, not wrapped around someone else’s API, and not on a vendor’s cloud and pricing schedule. Openmonoagent.ai is a terminal-native AI coding agent running entirely on local LLMs, with zero API costs, zero telemetry, and full ownership. Startuphakk Security offers a free option that surfaces three vulnerabilities, with a full report for $50.The through-line of this story is not a lone midnight post. It is a booking desk, a launch clock, overlapping safety hits, a funded panic, a regulatory playbook that looks like the one Bankman-Freed tried, and a fight over whether software stays publishable. Open source is the answer given here to the compute asymmetry. Concentrating inference in one or two labs, with government as the gate, is the outcome to refuse.
