AI News, Sep 30, 2026: Google announces Gemini 4 Argon, limited to vetted cyber defenders at first

Google launches Gemini 4 Argon while the White House's voluntary AI safety deal and an FTC probe shape oversight.

1. 01
   Independent tests put Gemini 4 Argon level with GPT-6 Astra, behind ClaudeThe Decoder

   Independent testing shows Gemini 4 Argon matching OpenAI's GPT-6 Astra but trailing Anthropic's Claude Opus 5.5, while using more than twice as many tokens per task.

   Why it matters

   Argon's per-token price is low, but it uses over twice as many tokens per task, so real costs for users and businesses may be higher.
2. 02
   Google announces Gemini 4 Argon, limited to vetted cyber defenders at firstThe Verge AI

   Google announced Gemini 4 Argon, a model for software engineering, legal and finance work, and cybersecurity, with initial access restricted to trusted cyber defenders.

   Why it matters

   Google's newest AI model is not open to everyone yet; early access goes to select testers and security teams, so most people must wait.
3. 03
   ElevenLabs valuation doubles to $22 billion in employee share saleTechCrunch

   AI voice startup ElevenLabs doubled its valuation to $22 billion through a $300 million employee share sale co-led by Wellington and T. Rowe Price.

   Why it matters

   Investor money is flowing into AI voice tools, which suggests more voice features will appear in apps and services you use.
4. 04
   Google replaces Gems with reusable Skills in GeminiThe Decoder

   Google is replacing Gems with Skills in Gemini chat, reusable prompts invoked by typing "/", with Gems phased out starting in November and existing ones migrating automatically.

   Why it matters

   If you saved custom Gems in Gemini, they will move to the new Skills format automatically after the phase-out begins in November.
5. 05
   FTC opens investigation into OpenAI, Anthropic and other AI labsThe Decoder

   The FTC is formally investigating OpenAI, Anthropic and other leading AI labs over possible consumer protection violations, using binding demands for documents and executive testimony.

   Why it matters

   The probe could lead to rules or penalties affecting the chatbots and AI tools consumers use every day.
6. 06
   DoorDash launches an AI agent you can text to order foodTechCrunch

   DoorDash launched an AI agent that customers can text to order food, aiming to compete with Uber Eats and Grubhub.

   Why it matters

   You can now order meals by sending a text message instead of browsing the app.
7. 07
   Trump and tech CEOs sign a voluntary AI code of conductThe Decoder

   President Trump and leaders from companies including Meta, Nvidia and xAI signed a "morally binding" AI code of conduct, with outside auditors meant to check that models work as intended.

   Why it matters

   The deal sets safety promises for major AI companies, but it is not legally enforceable, so protections for users remain unclear.