Busy weekend in AI land. The White House stood up a task force with a name that sounds like a Saturday morning cartoon, Google shut the door on a bug bounty because bots were flooding it with junk, Tencent found a legal side door to 100,000 American AI chips, and one of OpenAI's longest-tenured safety people walked out and wrote about it in The Atlantic. Let's go.
01 of 04
Trump Launches the "Super Intelligence Force"
On Sunday, President Trump announced the Super Intelligence Force on Truth Social. Director of National Intelligence Jay Clayton chairs it. The vice chairs are FTC Chairman Andrew Ferguson, Office of Personnel Management Director Scott Kupor, and Pentagon CTO Emil Michael. Vice President JD Vance, Defense Secretary Pete Hegseth, and Treasury Secretary Scott Bessent also sit on it, with former Secretary of State Condoleezza Rice and venture capitalist David Sacks advising from the outside.
The charter says the group has to "ensure that America continues to lead the World in Super Intelligence" and plan for "SI-enabled threats" while "preventing overregulation and regulatory capture that would stifle innovation and competition." Clayton told The Wall Street Journal the group owes a report on risks and opportunities within 120 days, including recommendations on disclosing security breaches. Asked about calls to slow AI down, he said, "I don't think any American should think that that's a good strategy."
This caps off a wild two weeks. Last week Trump signed an executive order telling federal agencies to say "super intelligence" instead of "artificial intelligence" in official correspondence. And on September 29, six CEOs signed a voluntary White House pledge called the Joint Commitment on Frontier Responsibilities: Dario Amodei of Anthropic, Greg Brockman of OpenAI, Sundar Pichai of Google, Mark Zuckerberg of Meta, Elon Musk of xAI, and Jensen Huang of Nvidia. They promised robust model monitoring, internal teams to check the controls actually work, outside auditors, and regular meetings to set safety standards. None of it is binding.
Here is where I land. Renaming a thing does not change the thing. My tractor does not get more horsepower if I start calling it a race car. A nonbinding pledge where the companies pick their own auditors plus a task force whose charter is half about not regulating too much is a pretty clear signal of where this is headed. I will read the 120-day report when it lands, but I am not holding my breath for teeth.
Sources: SiliconANGLE, Al Jazeera on the pledge
02 of 04
Google Pauses Its Open Source Bug Bounty Because of AI Slop
As of October 1, Google stopped taking new product vulnerability submissions to its Open Source Software Vulnerability Rewards Program, the OSS VRP. That program paid anywhere from $100 to $31,337 for security holes in Google's open source projects like Golang, Angular, and Protocol Buffers. Google's reason, in its own words: "This pause is due to a significant rise in automated submissions, the vast majority of which are not valid."
Supply chain reports and anything already in the queue still get handled. Researchers can also still go through the Patch Rewards Program, which pays up to $15,000, or the Cloud VRP. Google says it will share updates in Q1 2027. It did not put a number on how many junk reports it got.
Google is not the first to tap out. The curl project ended its HackerOne bounty back in January 2026 after getting buried in AI-generated reports. For scale on what is at stake, Google paid out $17.1 million across its reward programs in 2025 to more than 700 researchers.
What gets me about this one is who pays the price. It is not the guy running a script that spits out fake vulnerabilities. It is the real researchers who just lost a paycheck, and the open source maintainers who had to read all that garbage. Cheap AI made it free to spam, and when spam is free the honest folks get crowded out. Somebody is going to have to figure out a cost to submit, a reputation gate, something. Otherwise every bounty program on earth ends up looking like this.
Sources: Google VRP on X, BleepingComputer
03 of 04
Tencent Rents 100,000 AI Chips From Oracle, Legally
The Financial Times reported Thursday that Tencent signed a five-year, roughly $7 billion deal to lease about 100,000 advanced Nvidia chips sitting in Oracle data centers across Southeast Asia. About 30% is paid upfront. Tencent plans to use the compute for its AI models and agent tools. The reports did not name the exact Nvidia chip or which countries the data centers are in.
Here is the kicker. Chinese companies cannot buy these chips directly because of U.S. export controls, but U.S. rules do not currently stop them from renting the compute overseas. Tencent already pulled something similar last December, getting at Nvidia Blackwell chips through a data center in Japan. And ByteDance's Singapore arm got access to 2,304 Nvidia B200 GPUs through UK provider Nscale's data center in Norway.
What I want to know is what the export controls are even for at this point. We tell China they cannot buy the truck, then rent them the truck by the hour, with gas. Oracle is not doing anything illegal, and I do not blame a company for taking $7 billion. But Washington either closes the cloud loophole or admits the chip rules are mostly theater. You cannot have it both ways.
Sources: TrendForce, Benzinga
04 of 04
An OpenAI Safety Lead Quits and Says the Culture Is Broken
David Robinson spent three and a half years at OpenAI, which makes him one of the longest-tenured people there. He helped draft the company's Preparedness Framework and led the safety reports that ship alongside its big launches, covering twelve frontier model launches. On Saturday The Atlantic ran his exit essay, "I Quit OpenAI Because Its Culture Is Broken."
His core argument is that OpenAI leans on "iterative deployment," shipping first and patching safeguards after problems show up, and that this approach guarantees "periodic failures" that get bigger as models get more capable. "The time for trial and error is over," he wrote. He thinks frontier labs should run like nuclear plants or busy airports, with layers of redundancy and slow, careful planning, and noted he never met a colleague with experience keeping reactors from melting down or airplanes in the air. OpenAI spokesperson Drew Pusateri pushed back, saying the company is "making sure our models don't become more capable than we can safely manage and secure."
Robinson even called himself "something of a cliché," and he is not wrong. We have seen this exit essay before. But the details matter. This is not some random engineer. This is the guy who wrote the safety reports. When the person signing off on the homework says the grading is broken, I take that more seriously than any pledge signed at the White House the same week.
Sources: TechCrunch, Calcalist