B12 Magazine

Daily from the four markets B12 Group works in

Technology

AI that attacks on its own: what Google's new warning means for every business

Google's threat researchers say attackers now use AI agents that turn a fresh security patch into a working exploit without a human in the loop.

NewsB12 Group4 min read

Google's threat intelligence group has reported a shift that security teams have long feared. Attackers are no longer only asking AI tools for help; some are running autonomous, multi-agent systems that carry out several stages of an attack by themselves. In at least one operation, Google said, no human was needed to turn a newly released software patch into a working exploit.

Why the patch window is closing

When a vendor publishes a security fix, it also reveals, indirectly, where the weakness was. Attackers have always studied patches to build exploits for systems that have not yet been updated. That work used to take days or weeks of skilled effort. If an AI agent can do it in hours, the time a business has to apply updates safely shrinks dramatically.

A pattern, not a one-off

Google's findings arrive alongside other reports in the same week. Amazon's threat intelligence team described an AI agent campaign that searched for weak spots in internet-exposed AI applications, and the FBI has highlighted growing abuse of AI-driven penetration-testing frameworks. Investors have noticed too: AI security start-ups continue to raise money at high valuations.

What to do now

For most organisations the response is not exotic. It is discipline, applied faster:

Patch quickly. Critical updates for internet-facing systems, such as websites, online stores, VPNs and email, should be measured in days, not weeks.

Know what is exposed. Keep an up-to-date list of every server, domain and application reachable from the internet, including the forgotten test sites.

Watch for the unusual. Monitoring that alerts on new logins, new admin accounts and unexpected traffic catches what prevention misses.

Test before attackers do. Regular security audits and penetration tests find the weaknesses that an automated attacker would find first.

The takeaway

AI is making attacks cheaper and faster. The businesses that stay safe will be those that treat updates, visibility and testing as routine, run by people who know their systems. That is the daily work of B12 Group IT Solutions & Services.

More from Technology

Call us free