Technology

Autonomous AI Agents Are Now Attacking Networks

Screens in a security operations centre.
Written by Abigail Allen

Security researchers at Google’s Threat Intelligence Group and Mandiant say attackers are now operationalising autonomous multi-agent frameworks,.

Security researchers at Google’s Threat Intelligence Group and Mandiant say attackers are now operationalising autonomous multi-agent frameworks, automated vulnerability-exploitation pipelines and large-language-model hijacking to carry out intrusions at machine speed. The findings, published on November 16, mark a significant threshold: instead of a human operator directing each step of an attack, coordinated AI agents can compress timelines and execute multi-stage intrusions with far less human involvement. Defenders have warned about this possibility for years, and the new report suggests it has moved from theory to practice. Organisations that still assume attacks unfold at human pace may find their response times are no longer enough.

What the Evidence Shows

The shift changes the economics of both attack and defence. Automated agents can probe thousands of systems, test exploits and adapt to defences far faster than a person, which means the window between a vulnerability being disclosed and being exploited keeps shrinking. Security teams are responding with their own AI tools for detection and response, but the balance is uncertain. The practical advice is familiar yet more urgent: patch quickly, limit access, monitor continuously and assume that intrusions can escalate in minutes rather than days. The era of AI-versus-AI security has arrived, and the organisations that prepare for it now will be far better placed than those that wait. Small businesses are not exempt from this shift, since automated attackers can scan the internet for any exposed system, and experts say the same defences that protect large companies, such as strong passwords, two-factor authentication and regular updates, now matter more than ever for organisations of every size. Reporting and analysis from sources such as MIT Technology Review track the same developments, and specialists following autonomous AI agents attacking computer networks say the direction is now consistent across independent lines of evidence.

The Wider Context

The wider context is an industry learning to ship capability and restraint together. Products built around autonomous AI agents attacking computer networks now launch with safety evaluations, usage data and rollback plans as standard, because enterprise buyers and regulators ask for them in the first meeting. Companies that treated those questions as friction are discovering they have become the purchase criteria.

What Happens Next

What happens next for autonomous AI agents attacking computer networks depends less on another breakthrough than on the unglamorous work of reliability, pricing and support. The buyers evaluating it now are not enthusiasts; they are operators with budgets and auditors, and their checklists will decide the timeline. On the evidence of the current pilots, the technology has earned the right to be judged by them. You can follow related developments in our coverage of technology.

There is a practical lesson in all of this for readers deciding what to do today. Start with the verified facts, be wary of anyone promising certainty, and prefer reversible steps while the evidence is still developing. Circumstances differ by country, income and personal situation, so general guidance should always be weighed against your own context before you act on it in any significant or lasting way.

About the author

Abigail Allen

Leave a Comment