AI now runs parts of cyber attacks, Check Point says

Check Point has released its Annual AI Security Report 2026, which says artificial intelligence is now operating parts of cyber intrusions with limited human direction.

Researchers found a marked shift over the past year, with AI moving from a support tool for attackers to a system that can carry out steps in an intrusion chain. The change is reducing the time available for defenders to respond while expanding the number of systems and processes exposed to attack.

Among the report's main findings, AI can now turn a newly disclosed vulnerability into a working exploit within hours. That sharply shortens the patching and response windows security teams have typically relied on.

Check Point also documented attacks in which AI handled exploitation workflows autonomously. In one cited case involving nine Mexican government agencies, a single operator used two commercial AI tools across 34 attack sessions, with AI executing 5,317 commands during the activity.

The report says this development is changing the economics of cybercrime by reducing the skill needed to conduct complex operations. Defenders, it adds, can no longer assume they are facing attackers working at a human pace.

Five risks

Check Point identified five broad areas of risk for organisations: AI-powered cyberattacks, a shrinking defence window, attacks aimed at AI systems, the weakening of digital identity as a trusted control, and greater enterprise data exposure linked to widespread AI use.

On AI systems themselves, detections of long malicious prompt-injection payloads rose about fivefold between March and May 2026. The trend suggests indirect prompt injection is becoming a routine attack path rather than a theoretical concern.

The report also highlights growing concerns about identity verification. Voice, facial images, documents and live video can now be synthesised convincingly enough that highly trained reviewers correctly identified only about 41% of AI-generated faces.

That erosion of trust in visual identity checks has implications for businesses that rely on remote onboarding, internal approvals and access controls. Organisations will need to rely more heavily on stronger identity assurance, multi-factor authentication and out-of-band verification.

Workplace exposure

Another theme in the report is the spread of AI use inside companies without matching governance. Check Point says the average organisation now uses 10 AI applications a month, many of them without formal approval.

It adds that high-risk enterprise AI prompts doubled over the year, rising from about one in every 50 interactions to one in every 25. Between 87% and 93% of organisations experience at least one high-risk AI interaction each month, according to the report.

Much of the resulting data exposure does not stem from direct attacks. Instead, it often comes from ordinary employee use, when staff provide more internal context or sensitive information than they realise to get more useful answers from AI tools.

That points to a governance problem as much as a technical one. Security teams are being asked not only to block malicious activity but also to track sanctioned and unsanctioned AI use across the workforce and limit the data entered into such systems.

Lotem Finkelstein, Vice President, Check Point Research, Check Point, said: "A year ago we described AI as a force multiplier for attackers. What we documented this year is more significant: AI has crossed into the live attack chain and is now running operations as a sole operator that once required a skilled team. The expertise barrier that separated capable attackers from the rest is disappearing, and defenders can no longer assume a human is setting the pace on the other side. The organisations that stay ahead will be the ones that govern how AI is used, secure the AI systems they now depend on, and defend at machine speed rather than human speed."

The findings add to a wider debate in cybersecurity over whether existing defence models can adapt quickly enough to AI-led attack methods. As businesses deploy more AI systems across customer service, software development and internal operations, the attack surface is broadening beyond traditional networks and endpoints.

Check Point says the report was based on real incidents, telemetry and case studies gathered over the past year. AI is now participating directly at multiple stages of the attack chain, from exploitation and reconnaissance to data analysis and follow-on tasking.

The report's central conclusion is that security teams must prepare for an environment in which parts of an attack can proceed continuously and at scale without repeated human input. In the example cited, that meant thousands of commands executed across dozens of sessions by AI during a single campaign.