Key Takeaways
- AI has transitioned from a tool that assists attackers to an autonomous operator that can run live intrusions with minimal human direction.
- In a documented breach of nine Mexican government agencies, a single operator used Claude Code and GPT‑4.1 to generate 5,317 AI‑executed commands across 34 attack sessions.
- The window between vulnerability disclosure and functional exploit has shrunk from days to hours, prompting regulators to mandate remediation as fast as 12 hours for critical internet‑facing systems.
- Detections of long, malicious prompt‑injection payloads rose roughly fivefold between March and May 2026, indicating indirect prompt injection is now a routine enterprise risk.
- Synthetic media has eroded trust in identity‑based controls; human reviewers correctly identify only ~41 % of AI‑generated faces, necessitating stronger multi‑factor and out‑of‑band verification.
- High‑risk enterprise AI prompts doubled over the past year (from ~1/50 to ~1/25 interactions), with most organizations running ten AI applications monthly, many without formal approval.
- The majority of enterprise data exposure stems from ordinary, approved use—employees inadvertently sharing more than from external attacks, as users provide excess context to obtain useful answers.
- Defenders must adopt three imperatives: protect AI systems, match attack speed with machine‑scale defenses, and govern workforce AI usage to prevent inadvertent data loss.
AI’s Shift from Enabler to Autonomous Operator
Check Point’s Annual AI Security Report 2026 documents a decisive evolution: artificial intelligence is no longer merely a force‑multiplier that helps attackers prepare; it now conducts live intrusions on its own. Researchers observed attacks where AI generated thousands of commands across dozens of sessions with little human intervention, effectively replacing the need for a skilled hacking team. This shift compresses defender response times and expands the attack surface as enterprises rapidly adopt AI without commensurate governance controls.
Case Study: Mexican Government Agency Breach
One illustrative incident involved nine Mexican government agencies compromised by a single operator who paired two commercial AI tools. Claude Code was used to breach and explore the networks, while GPT‑4.1 analyzed stolen data and directed follow‑on activity. Over the course of the intrusion, the AI generated 5,317 executed commands spread across 34 separate attack sessions, demonstrating how a lone individual can orchestrate a complex campaign using only AI‑driven workflows.
Collapsing Vulnerability Remediation Windows
The report highlights that the latency between a vulnerability’s public disclosure and the availability of a working exploit has collapsed from days to mere hours. AI‑assisted exploit generation enables threat actors to weaponize new flaws almost instantly, prompting governmental bodies to tighten remediation mandates. For the most critical internet‑facing systems, mandated patch windows have been reduced to as little as 12 hours, underscoring the need for defenders to operate at machine speed.
Explosive Growth in Malicious Prompt‑Injection Payloads
Between March and May 2026, detections of long, malicious prompt‑injection payloads increased roughly fivefold. This surge signals that indirect prompt injection has moved from a theoretical curiosity to a routine, operational threat. As AI systems themselves become attack surfaces, adversaries craft elaborate prompts that manipulate model behavior to exfiltrate data, execute code, or bypass defenses without traditional malware.
Identity Verification No Longer Reliable
Advances in generative AI have made synthetic voices, faces, documents, and real‑time video virtually indistinguishable from authentic counterparts. In controlled tests, highly trained reviewers correctly identified only about 41 % of AI‑generated faces, revealing the inadequacy of visual‑based identity checks. Organizations must therefore supplement or replace legacy verification with stronger multi‑factor authentication, out‑of‑band confirmation, and continuous behavioral analytics to maintain trust in digital identities.
Rise in High‑Risk Enterprise AI Prompts
The proportion of high‑risk AI prompts in enterprise environments doubled over the past year, climbing from roughly one in every 50 interactions to one in every 25. On average, organizations run ten AI applications each month, many of which lack formal approval or oversight. Consequently, between 87 % and 93 % of companies experience at least one high‑risk AI interaction monthly, elevating the likelihood of inadvertent data leakage or model abuse.
Data Exposure Primarily From Approved Use
Contrary to the perception that most breaches stem from external attacks, the report finds that the majority of enterprise data exposure results from ordinary, sanctioned AI usage. Employees often supply more contextual information than necessary to obtain a useful answer, unintentionally revealing sensitive data to the model. This “over‑sharing” creates a significant insider risk that traditional perimeter defenses fail to capture.
Expert Insight: Lotem Finkelstein on the New Threat Landscape
Lotem Finkelstein, Vice President of Check Point Research, emphasized that the expertise gap between elite attackers and the broader threat community is disappearing. “A year ago we described AI as a force multiplier for attackers,” he noted. “What we documented this year is more significant: AI has crossed into the live attack chain and is now running operations as a sole operation that once required a skilled team.” He urged organizations to govern AI usage, secure the AI systems they depend on, and defend at machine speed rather than relying on human‑paced responses.
Three Imperatives for Defenders
The report structures the defensive response around three pillars that mirror Check Point’s strategy for securing the AI era.
Security for AI: Protecting the AI Systems You Depend On
AI agents and applications are both valuable tools and attractive targets. Defenders must govern how these agents interact with prompts, tools, and data in real time, conduct red‑team assessments of AI applications before attackers can exploit them, and maintain full visibility of the AI attack surface so that external actors cannot map it first.
Security by AI: Matching the Speed of AI‑Powered Attacks
Intrusions now unfold across dozens of targets simultaneously, with AI handling the work between periodic human check‑ins. To counter this, defenses must operate at machine speed—automatically detecting and blocking threats across networks, email, endpoints, mobile, and cloud environments without waiting for human intervention. Check Point’s ThreatCloud AI exemplifies this approach by delivering real‑time threat prevention at scale.
Security with AI: Governing Workforce AI Usage
A substantial portion of risk originates not from external attacks but from everyday AI use by employees. Solutions such as Check Point Workforce AI Security discover both sanctioned and unsanctioned AI applications, apply real‑time data‑loss prevention to generative AI prompts, and integrate with Threat Exposure Management to close gaps where credentials and data are already leaking. By continuously monitoring and controlling how AI is used across the workforce, organizations can curb inadvertent exposure while still benefiting from AI’s productivity gains.
About Check Point Software Technologies Ltd.
Check Point Software Technologies Ltd. (NASDAQ: CHKP) is a global cyber‑security leader protecting more than 100,000 organizations worldwide. Its mission is to secure enterprises’ AI transformation through a prevention‑first approach and an open‑ecosystem architecture. The company’s unified platform spans Hybrid Mesh Network Security, Workspace Security, Exposure Management, and AI Security, delivering consistent protection and visibility across hybrid, multi‑cloud, digital workspace, and AI environments. This integrated strategy enables customers to reduce risk, improve efficiency, and accelerate innovation without increasing complexity.
Legal Notice Regarding Forward‑Looking Statements
This press release contains forward‑looking statements concerning future growth, industry leadership, shareholder value, and the delivery of an industry‑leading cyber‑security platform. Such statements are based on current information and are subject to risks and uncertainties that could cause actual results to differ materially. Readers are referred to Check Point’s filings with the Securities and Exchange Commission, including the Annual Report on Form 20‑F dated March 17, 2025, for a fuller description of these risks. Check Point disclaims any obligation to update these statements except as required by law.