Otto365 AI Brief #034

Gemini: Google builds its new model for cyber defenders

Google's new Gemini built for security work

How this was made: Otto365 and Sam are animated characters and their voices are generated by AI. Sam is a fictional small-business owner. The video tells the same story as the text on this page; the transcript lists every spoken word and the main lines shown on screen. Spotted an error?
  • Published
  • By the Otto365 AI Brief. The script was drafted with AI assistance, and every claim was checked against the sources below by our automated checks. Published on under the standing approval of Peter Ferguson, Editor.
  • Announced 28 Oct 2026
  • Voiced by AI

What happened

Google's new Gemini model is built for security work, not for you yet. Google DeepMind announced Gemini 4 Carbon, its next frontier model, on the 28th of October 2026. Google DeepMind calls it its strongest model, focused on cybersecurity defence and long tasks.12

Three things to know

  1. Finds, checks, patches First, security. Google DeepMind says Carbon finds real vulnerabilities, checks which ones can be exploited, and patches the root cause with working code fixes. It scores 88.2% on SEC-bench Pro for finding flaws.1
  2. Runs for days on bugs Second, Google DeepMind says its teams have seen Carbon run for days to patch subtle bugs across large codebases. It says Carbon resisted every evaluated prompt injection attack and roughly halves unsupported claims versus Argon.1
  3. Defenders: end Oct 2026 Third, access. A subset of cyber defenders gets Carbon first, then trusted partners via the Fairwind Program; developers, businesses and consumers have no date yet. Pricing matches Argon: 2 dollars per million input tokens, 10 per million output.12

Why it matters to you

Otto365's take (opinion)

So, my take. Frontier AI is being built for defenders before everyone else, and a UK small business meets it through its IT provider, not a chatbot.1

Do this week

Try this: ask your IT provider how AI helps them patch weaknesses.

Sources

  1. Google DeepMind: Introducing Gemini 4 Carbon (opens in a new tab) Primary source, accessed 29 Oct 2026
  2. The Verge: Google launches Gemini 4 Carbon with a ‘significant leap’ in cybersecurity (opens in a new tab) News report, accessed 29 Oct 2026
The quotes we checked (17)

  • Today, we're announcing the next frontier model: Gemini 4 Carbon, our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

    Google DeepMind · source 1

  • Gemini 4 Carbon: our next frontier model Oct 28, 2026

    Google DeepMind · source 1

  • Oct 28, 2026, 8:03 PM UTC

    The Verge · source 2

  • Gemini 4 Carbon is our strongest model with state-of-the-art cybersecurity and long horizon capabilities.

    Google DeepMind · source 1

  • It achieves state-of-the-art performance across the cybersecurity defense lifecycle - vulnerability detection, validation, and remediation.

    Google DeepMind · source 1

  • Carbon achieves a state-of-the-art 88.2% on SEC-bench Pro, autonomously uncovering complex flaws buried in real-world software.

    Google DeepMind · source 1

  • cyber defenders increasingly rely on models to check which vulnerabilities are exploitable.

    Google DeepMind · source 1

  • Carbon reaches 71.0% on Collinear's CWE-bench, turning validated findings directly into working code fixes.

    Google DeepMind · source 1

  • Internally, our teams have seen Carbon run for days to persist through dead ends to tackle complex, multi-step engineering tasks like root-causing and patching subtle bugs across large codebases and low-level systems.

    Google DeepMind · source 1

  • While Argon previously led the Gray Swan's Indirect Prompt Injection (IPI) Benchmark in prompt injection resistance with a very low 0.7% attack success rate, Carbon achieves 0.0% - resisting every evaluated attack attempt.

    Google DeepMind · source 1

  • In our internal testing, Carbon cuts unsupported claims by approximately 2x over Argon.

    Google DeepMind · source 1

  • Starting this week, we will make Carbon available to a subset of cyber defenders and will gradually phase access to additional trusted partners through our Fairwind Program.

    Google DeepMind · source 1

  • We're continuing to gather feedback and iterate on guardrails before making Carbon available to developers, businesses, and consumers via our product surfaces as soon as possible.

    Google DeepMind · source 1

  • It's currently only available to a select group of cybersecurity defense companies, but it will eventually expand to other partners in its Fairwind Program.

    The Verge · source 2

  • as Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95% off input token price.

    Google DeepMind · source 1

  • It will launch at the same competitive introductory price as Argon, at $2 per million input tokens and $10 per million output tokens.

    The Verge · source 2

  • we look forward to bringing Carbon to developers, businesses, and consumers soon.

    Google DeepMind · source 1

Transcript

Full transcript

Sam: Otto! Google's new Gemini, built for security work!

Play from 0:03 Otto365: I know. Google DeepMind announced Gemini 4 Carbon on the 28th of October. Sam: What does it actually do? Otto365: Google DeepMind calls it its strongest model. Focused on cybersecurity defence and long tasks.

Play from 0:18 Sam: Defence how, exactly? Otto365: Google DeepMind says it finds real vulnerabilities and checks which can be exploited. Then it patches the root cause with working code fixes. Sam: Any good? Otto365: It scores 88 point 2% on SEC-bench Pro for finding flaws.

Play from 0:33 Sam: What about the long tasks? Otto365: Google DeepMind says its teams have seen Carbon run for days. It patches subtle bugs across large codebases. Sam: Is it safe? Otto365: It says Carbon resisted every evaluated prompt injection attack. And it roughly halves unsupported claims versus Argon.

Play from 0:50 Sam: When can I have it? Otto365: A subset of cyber defenders gets it first, then trusted partners via the Fairwind Program. Developers, businesses and consumers have no date yet. Sam: And the price? Otto365: Pricing matches Argon. $2 per million input tokens, $10 per million output.

Play from 1:07 Sam: So it can guard my website! Otto365: My take: frontier AI is being built for defenders before everyone else. A UK small business meets it through its IT provider, not a chatbot. Sam: I'd already cleared a desk for it.

Play from 1:22 Otto365: Keep the desk. Try this: ask your IT provider how AI helps them patch weaknesses.

Otto365: I'm Otto365.

Follow the Brief.

On screen

Play from 0:00
Otto365 AI Brief #034, Google Gemini, 28 Oct 2026
Google's new Gemini built for security work
AI built for defenders
Play from 0:03
Carbon: built for defence
Announced 28 Oct 2026
Gemini 4 Carbon
Cybersecurity defence
Long-horizon tasks
Google's strongest model
Source: Google DeepMind, 28 Oct 2026
Play from 0:18
Finds, checks, patches
Find the flaw
Prove exploitable
Patch root cause
Working fix
88.2% SEC-bench Pro: finds real flaws
Play from 0:33
Runs for days on bugs
Large codebase
Runs for days
Past dead ends
Bug patched
Resists every injection test
Halves unsupported claims
Play from 0:50
Defenders: end Oct 2026
Cyber defenders
Fairwind partners
Developers, businesses
Input: $2 per million, Same as Argon
Output: $10 per million
Play from 1:07
AI built for defenders
Chatbot: Answers questions
Defence model: Finds, fixes flaws; Runs for days
Your business
IT provider
AI finds flaws
Patched
Play from 1:22
Ask IT about AI patching
Voiced by AI. Fact-checked against its sources. Apps 365 sells AI and Microsoft 365 services. Not endorsed by companies named.

Voiced by AI. Fact-checked against its sources. Apps 365 sells AI and Microsoft 365 services. Not endorsed by companies named. Report an error.

  1. #014

    Gemini: Google's Argon model is no longer gated

    Google AI announced Gemini 4 Argon for paid users and developers on the 13th of October 2026. This week: Draft one proposal, check it.

  2. #003

    Gemini: Google's new model targets legal and finance work, with a catch

    Google DeepMind announced Gemini 4 Argon, its new frontier model, on the 30th of September 2026. This week: List one multi-step task.

  3. #009

    ChatGPT: real cartoonists' names on fake New Yorker cartoons

    Nieman Lab reported on the 5th of October 2026 that ChatGPT is adding real New Yorker cartoonists' signatures to AI-generated cartoons. This week: Check images for signatures.