Dark low-key editorial still-life of a steel clipboard stack with five clipped slate briefing sheets, amber binder clip accent and cool rim light on a charcoal desk, no logos or faces.
AIMarketsNewsTech

The Thursday AI brief: Google ships Gemini 4 Argon Fairwind-first, the FTC probes frontier labs, and Trump renames AI to Super Intelligence

Aaron Harme

Wednesday ran from a Super Intelligence rename in the morning to an FTC probe of OpenAI and Anthropic at noon, then Gemini 4 Argon for vetted cyber defenders in the evening. OpenAI’s Moonshot distillation disruption and Reddit’s RSS-and-API shutdown closed the pipes.

The five items below carry a live link: Argon, the FTC probe, the SI rename, the Moonshot campaign, and Reddit’s free-feed kill. Meta’s research tax credits, Google’s publisher-pay pilot, DoorDash text ordering, Altman’s IPO-until-safe line, GLM-5.3, and Oura’s IPO pause sit in the notes.

In this Brief

  • Google shipped Gemini 4 Argon first to trusted cyber defenders through Fairwind, with a 1M-token output window
  • The FTC confirmed a sweeping probe of OpenAI, Anthropic, and other frontier labs, with METR also expected as a target
  • Trump ordered federal agencies to say Super Intelligence instead of AI in official settings
  • OpenAI says people tied to China’s Moonshot ran a mass distillation campaign on its models
  • Reddit is killing RSS in November and the public API by March 2027 over AI bots

Google just shipped Gemini 4 Argon, Fairwind defenders first

Setup: Google announces Gemini 4 Argon, its next frontier model, the company’s most capable general AI system, rolling out first through the Fairwind Program, Google’s vetted early-access path for trusted cyber defenders. Koray Kavukcuoglu’s post frames Argon for complex, long-horizon work across software engineering, legal and finance knowledge tasks, and cybersecurity defense. Introductory API pricing is $2 per million input tokens and $10 per million output tokens, with cached input at 95% off. The output window jumps to 1 million tokens, up from a prior 64K ceiling.

Axios reports Argon outperforms GPT-6 Astra on several coding and knowledge-work marks after roughly a year since Gemini 3. 9to5Google prints the DeepSWE score: 77.9% against Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%, and says after the introductory period the price moves to $4 and $20 per million input and output tokens.

For trusted defenders and internal teams, Google is releasing Argon without cyber guardrails so they can use full frontier cyber-defense capabilities. Wiz is already using it via Scan for Good; an early demo uncovered a critical vulnerability exposing sensitive personal information across healthcare software used by hospitals worldwide that previous frontier models had missed. Broader release is framed for paid API customers and Google AI Ultra subscribers after more testing.

Tulsee Doshi, head of Gemini products, calls Argon a well-rounded model with frontier capabilities across several domains.

See full breakdown at @Aaron_Harme on X.

Longer cut: Google Gemini 4 Argon with a 1 million token output window.

The FTC just opened a sweeping probe of OpenAI and Anthropic

Setup: The New York Post reports a sweeping FTC probe of Anthropic, OpenAI, and other frontier labs over potential dangers their technology poses to consumers. A senior FTC official told the Post that Chairman Andrew Ferguson opened the work a few weeks ago. The agency is drafting civil investigative demands, formal orders that work like subpoenas for documents and testimony, to compel executives to talk about their products and the dangers those products may pose to consumers. CIDs are expected in the coming weeks. METR, a Berkeley AI watchdog, is also expected to be a target.

CBS News confirms the investigation went public Wednesday and says the agency first opened it this summer. The Post says Ferguson launched it before the so-called Hugging Face incident. An official told the Post the agency is in the investigative phase and not telling firms to stop. Anthropic and OpenAI did not immediately respond.

On Fox News September 20, Ferguson warned against letting firms whip everyone into a panic and then demand regulations they can clear as a competitive moat, arguing competition is what keeps America ahead of China on AI.

See full breakdown at @Aaron_Harme on X.

Longer cut: FTC sweeping probe of OpenAI, Anthropic, and frontier labs.

Trump just ordered federal agencies to say Super Intelligence instead of AI

Setup: The Verge reports Trump signed an executive order directing the US executive branch to refer only to Super Intelligence in official policy websites, policy documents, and press releases. An executive order is a presidential directive to federal agencies. Past regulations and older documents stay as written. At Tuesday’s America.gov launch, Trump called super the best word of all and the simplest, said Chinese President Xi Jinping loves it too, and compared artificial to fake news.

Gizmodo prints the order’s line that the executive branch will not acknowledge the usage of Artificial Intelligence and AI in any applicable setting. In brief remarks after the lunch, Trump said the name had been officially changed to SI, pronounced sigh, with a nod to Nvidia’s Jensen Huang. The order defines Super Intelligence via the technologies already covered by artificial intelligence in 15 U.S.C. § 9401(3).

Meta Public Affairs told Gizmodo that Mark has been using superintelligence publicly. Anthropic, OpenAI, SpaceX, Nvidia, and Google did not reply to Gizmodo in time.

Michael Kratsios, Assistant to the President for Science and Technology, has 60 days to propose legislation that could modify, expand, or supersede older AI terms.

See full breakdown at @Aaron_Harme on X.

Longer cut: Trump orders federal agencies Super Intelligence instead of AI.

OpenAI says people tied to China’s Moonshot ran a mass distillation campaign

Setup: OpenAI says it disrupted a coordinated campaign to extract protected reasoning, the model’s hidden scratch work while solving a problem, from its systems. Activity was consistent with adversarial distillation: systematic unauthorized use of one model’s outputs or reasoning to help train or improve another. Operators did not break encryption or compromise a database. They manipulated model interactions so protected reasoning could be reproduced at scale, violating terms of service.

Earliest observed activity landed in the first week of July. High-volume spikes on July 24 and 25 hit 16,000 requests using a relevant extraction pattern from over 4,000 users, with related prompt-pattern activity across a cluster of more than 15,000 users. OpenAI says it fully disrupted the campaign by July 28. The company attributes a core cluster to individuals associated with Moonshot AI, the lab behind Kimi. It banned fraudulent accounts, closed a pathway that allowed replay of another user’s encrypted reasoning, and shared the technique with the Frontier Model Forum and government channels.

OpenAI’s figures describe attempted extractions, and the page leaves open whether every operator was a single actor.

See full breakdown at @Aaron_Harme on X.

Reddit is killing RSS and the public API because of AI bots

Setup: TechCrunch reports Reddit is ending RSS feeds and public API access, blaming large-scale scraping and automated abuse. RSS is a free subscribe feed of new posts. The public API is the official machine interface outsiders used to read Reddit at scale. RSS support ceases November 13. Public API access ends by March 2027. After shutdown, products that leaned on programmatic access will need commercial deals for Reddit data.

Q2 “other revenue” beyond advertising grew 24% year over year to $43 million, the AI-licensing color TechCrunch prints. The Verge covers Old Reddit cutbacks in the same Wednesday package: logged-in access limited for scrapers, and in the next few months Old Reddit for logged-in users who used it in the last six months. Developers of approved third-party apps must register before January 12, 2027 or lose API access.

The forum that licensed its posts to AI labs is shutting the free pipes those bots still use. Moderators who leaned on RSS for alerts are steered to a Discord Relay Devvit app.

See full breakdown at @Aaron_Harme on X.

Also noted

  • Quartz reports Meta, citing the New York Times, claimed research tax credits by treating AI data centers as experimental. Savings rose from $700 million in 2023 to $2 billion in 2024 and $3.9 billion in 2025; the company’s own filings warn the IRS may claw them back.
  • The Verge reports Google’s publisher-pay pilot for AI Overviews, AI Mode, and Gemini: about 100 publishers, with one early entrant over $1 million in a year and another around $50,000 to $60,000.
  • The Verge reports DoorDash text ordering: an Ask DoorDash waitlist that builds the cart and checks out inside the thread, plus a northern California drone pilot and $7.99 retail returns expanding in November.
  • The Verge reports Altman saying OpenAI will not go public until it can make confident safety claims, with no firm timeline, and that waiting forever would also be bad for the world.
  • Anthropic’s GLM-5.3 research says Zhipu’s open-weight model can build end-to-end cyber exploits with safeguards attackers bypassed 64% to 100% of the time in Anthropic’s simulated tests.
  • The Wall Street Journal reports Oura postponed its IPO despite strong demand, citing uncertainty in the IPO market.

Sources

Discover more from Bitter Fool

Subscribe to get the latest posts sent to your email.

Leave a Reply

Your email address will not be published. Required fields are marked *