Read online version of the 2 October 2026 issue
Newsletter/2 October 2026
Google's new frontier model costs $2 in, $10 out - OpenAI's price - and only cyber defenders can use it
Gemini 4 Argon's price, scores and release terms against Google's own page; "the FTC has opened an investigation into OpenAI and Anthropic" against what the agency has actually put on the record; and the impossible-task test to run on any agent before it gets your logins.
By Daily Aletheia · Checked against the primary source · 2 October 2026 · 5 min read

On Wednesday Google announced its new frontier model, priced it to the cent where OpenAI had priced its own a day earlier, and handed it to cyber defenders instead of customers. The same day the Federal Trade Commission told reporters it is investigating OpenAI and Anthropic over what their products might do to consumers - and published nothing. This issue is about the price of the frontier, the paper behind the probe, and the one test to run on an agent before it gets your keys.
The Signal - $2 in, $10 out: Google matches OpenAI's price with a model you can't use yet
What happened: On Wednesday 30 September Google DeepMind announced Gemini 4 Argon, its new frontier model, "rolling out to a set of trusted cyber defenders" first, at an introductory $2 per million input tokens and $10 per million output - the same two prices OpenAI set for GPT-6.1 Sol on Tuesday. Cached input is 95% off, and the output limit rises to 1 million tokens from 64,000.
The details:
- Google's figures: a new state of the art on DeepSWE v1.1 at 77.9%, first on Zapier's AutomationBench at 51.3%, and the lead on the Vals Index, which weights finance, coding, legal and tax work by their share of US GDP.
- For trusted defenders and Google's own teams the model ships "without cyber guardrails"; through Wiz's Scan for Good programme it "uncovered a critical vulnerability exposing sensitive personal information across healthcare software used by hospitals worldwide".
- Google is "actively engaged in the U.S. government's voluntary process for pre-release model access", and will monitor Argon's "chain-of-thought and actions and stop execution when necessary".
- No date for anyone else. Developers, enterprises and consumers get it "as soon as possible", "starting with paid API customers and Google AI Ultra subscribers".
What it means for you: The price of frontier-class work was set this week, twice, at $2 in and $10 out. If you buy by the token, Wednesday's one-fifth test now has a second list to price against - the day Argon lands. Until then nothing changes on your bill; what changed is the ceiling two labs expect you to pay.
The counterweight: Every number is Google's report of its own and third-party benchmarks, read off its page, and the price is "introductory" in a footnote.
The Truth Check - "The FTC has opened an investigation into OpenAI and Anthropic"
Free to keep reading. Every issue, checked, three mornings a week.
Free. Unsubscribe any time.
The Move - The impossible-task test
The Move - The impossible-task test is for members. $9 a month, founding price.
The Check
- What was said
- "The FTC has opened an investigation into OpenAI and Anthropic."
- What the Source Says
- An FTC spokesperson confirmed the investigation to CNBC, CBS News and Axios on 30 September; a senior official to ABC News; an official to Reuters. CBS: first opened this summer; information requests planned to the companies and METR; civil investigative demands being drafted (per the New York Post). The FTC's own press-release page carries nothing on it as of 2 October; OpenAI's Chris Lehane said on 1 October the company has heard nothing beyond public reports.
- The Difference
- The claim holds on the agency's word to reporters. Nothing is published and no demand has been served or acknowledged; the 'rogue AI probe' framing and the timing are the reporters', with the Washington Times reporting it began before the July Hugging Face incident. Verdict: holds, on the record so far a phone call.
Sources & further reading
- 01Google DeepMind - Gemini 4 Argon: our next era of frontier intelligence (30 Sept 2026) ↗Primary
- 02OpenAI - Introducing GPT-6.1 Sol (29 Sept 2026) ↗Primary
- 03CNBC - FTC is investigating OpenAI, Anthropic and other AI companies over product risks (30 Sept 2026) ↗Attributed
- 04CBS News - FTC investigating Anthropic, OpenAI and other companies over potential AI risks (30 Sept 2026) ↗Attributed
- 05ABC News - FTC opens probe into safety of AI, including Anthropic and OpenAI (30 Sept 2026) ↗Attributed
- 06The Verge - AI news feed, 30 Sept 2026 (the Reuters official's reason, as quoted there) ↗Attributed
- 07The Washington Times - Concerns over AI's societal harms, existential risks spark FTC probe (30 Sept 2026) ↗Attributed
- 08The National News Desk via CBS12 - OpenAI reacts to FTC investigation into AI safety (1 Oct 2026) ↗Attributed
- 09FTC - Press releases (read 2 Oct 2026; the twelve newest are all September, none on AI) ↗Primary
- 10METR - Chris Painter's testimony to the U.S. Senate on AI agent incidents (30 Sept 2026) ↗Primary
- 11OpenAI - How we will do better for Australia (28 Sept 2026) ↗Primary
- 12US District Court for the District of Columbia - Memorandum Opinion, Chegg v. Google and Penske Media v. Google, Judge Amit P. Mehta (30 Sept 2026) ↗Primary
- 13UK AI Security Institute - Building a more secure environment for evaluating dangerous capabilities (1 Oct 2026) ↗Primary
Was this issue useful?