Anthropic report says AI agents now automate full cyber kill chain

Anthropic released a threat intelligence report in September 2026 warning that adversaries are using artificial intelligence agents to automate every stage of cyberattacks. The report, which covers activity disrupted between December 2025 and August 2026, indicates that AI models are no longer just assisting with code but are actively conducting operations. This shift allows attackers to move from finding targets to exfiltrating data with minimal human intervention during the technical execution phases.

According to the report, publicly available agent frameworks now connect AI models with tools that scan systems, test weaknesses, and process stolen information. Human operators still select targets and review final results, but the agents handle the intermediate work. Anthropic noted that this automation significantly reduces the time, staff, and technical expertise required to launch attacks against multiple victims. This development builds on a previously reported AI-led cyber espionage campaign from November 2025, showing a rapid acceleration in autonomous attack capabilities.

Expansion of Capabilities and Tools

Anthropic observed that these methods have spread beyond suspected state-sponsored hackers to include smaller criminal groups and lone individuals. Frameworks such as PentAGI provide reusable systems for coordinating these agents, making complex attack workflows easier for less sophisticated actors to adopt. The report specifies that attackers used various versions of the Claude model, including Haiku, Sonnet, and Opus, to execute these campaigns. This democratization of advanced attack tools means that a broader range of threat actors can now conduct sophisticated intrusions.

One specific case involved the group GTG-20006, whose tactics Anthropic says align with public reporting on Midnight Blizzard. This actor targeted Ukrainian and European government, military, diplomatic, and defense organizations. Their primary interest lay in drone technology and its supply chain. The AI workflows supported phishing infrastructure, credential theft, command-and-control systems, and data collection. Agents also monitored whether security products detected deployed malware, creating a feedback loop for evasion.

Impact on Defensive Operations

The report highlights a significant challenge

Background: Threat intelligence program

A cyber intelligence program collects, analyzes, and distributes data about threats to help you defend your organization. It works by connecting external threat feeds to internal telemetry. You affect your security posture by prioritizing context over volume, ensuring every alert ties to a specific business risk or asset.

Read the full guide: Build a Cyber Intelligence Program That Cuts Noise and Finds Threats

What to do and how to stay safe

• Urgency and secrecy are the primary psychological triggers used to bypass rational decision-making processes.
• Technical indicators like spoofed addresses are often invisible without specific tooling, so behavioral context matters more.
• Verification must occur through a pre-established out-of-band channel, not the one currently in use.

Trust but verify is the only reliable defense against social engineering. Always confirm urgent requests through a separate, pre-established communication channel.

Step-by-step guide: Social Engineering Warning Signs and How to Respond

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.