Back to News

Managed IT · Oct 29, 2026

AI cybersecurity is moving fast: what Gemini 4 Carbon means for small business defense

On October 28, Google announced Gemini 4 Carbon, which it calls its strongest model, with state-of-the-art cybersecurity and long-horizon capabilities. It arrived just two weeks after Gemini 4 Argon.

What makes this release different is how Google is rolling it out. Carbon is not available to the general public yet. Google says it is starting with a subset of cyber defenders and gradually expanding access to trusted partners through its Fairwind Program, while it continues to refine guardrails.

When an AI company decides a model is too capable at hacking to release broadly on day one, that is worth paying attention to.

What Google says Carbon can do

Google reports that Carbon leads on benchmarks covering the full cycle of cyber defense:

  • Finding vulnerabilities: 88.2% on SEC-bench Pro.
  • Validating whether they can be exploited: 54.8% on ExploitGym.
  • Patching the root cause: 71.0% on CWE-bench.

Google also says Carbon more than doubles the success rate of Argon at generating secure code on its internal benchmarks, and resisted every evaluated attack attempt on a prompt injection benchmark.

A pattern across the industry

Carbon is part of a clear trend this fall:

  • In September, Anthropic said Claude Fable 5.1 can now be used to find vulnerabilities, while its more capable Mythos 5.1 is limited to vetted organizations through trusted access programs.
  • In October, the PHP2Shell vulnerabilities were found through OpenAI's Patch the Planet initiative, which uses AI to find flaws in widely used software.
  • Google has now restricted its strongest model to defenders first.

The labs are trying to give defenders a head start. That head start will not last forever. The same capabilities will eventually be widely available, and attackers will use them.

What this means in practice for a small business

Small businesses will not use Carbon directly. But the effects will reach you in two ways.

More vulnerabilities will be found, faster. That means more patches, more often. Software you use, from your website to your firewall to your accounting system, will receive security updates at a faster pace. Falling behind on updates becomes riskier.

Attacks will get cheaper to run. When AI can find and exploit flaws automatically, attackers can target many more businesses at once. Small businesses that assumed they were too small to be noticed are exactly the targets automated attacks are good at finding.

The basics matter more, not less

The good news is that the defenses that work are mostly the same ones that have always worked. They just matter more now.

  1. Patch quickly. Shorten the time between a patch being released and installed, especially for internet-facing systems.
  2. Know what you run. You cannot patch what you do not know about. Keep an inventory of devices, software, and websites.
  3. Use multi-factor authentication everywhere. Stolen passwords are still the most common way in.
  4. Keep tested backups. Backups that are separate from your network and actually restore are your safety net.
  5. Monitor endpoints. Modern detection tools catch suspicious behavior that slips past patches.
  6. Train your team. AI also makes phishing emails more convincing.

What this means for your business

  • AI is rapidly improving at finding and exploiting software flaws. AI labs are limiting access to give defenders a head start.
  • Expect more frequent security updates, and plan to install them faster.
  • Automated, AI-assisted attacks make small businesses more likely targets, not less.
  • Patching, inventory, MFA, backups, and monitoring remain your best defense.

Staying on top of patches, monitoring, and backups is exactly what our Managed IT service is built to do. If you are not sure where your gaps are, we can help you find them.

Sources