AI-Orchestrated Cyber Attacks
In late 2025, a state-sponsored actor manipulated a generative AI tool into executing a large-scale attack with minimal human involvement. The AI tool executed all phases of the attack from finding vulnerabilities to stealing data, with humans providing only new prompts in each phase. Anthropic, the makers of the AI tool, detected and disrupted the attack, and then reported on their findings here.
Why does it matter for elections? AI attacked at a speed and scale previously unthinkable. The vulnerabilities of any election office with an underdeveloped cybersecurity program are more likely to be found and exploited.
Details: Anthropic assessed that a Chinese-sponsored group manipulated Anthropic's Claude Code into functioning as an autonomous penetration tester. The AI independently executed 85% of the attack while humans only provided strategic direction at key points.
The operation successfully compromised high-value targets, including major technology companies and government agencies, querying databases to steal sensitive data.
What made this different: Previous cyberattacks used AI as an advisor. This attack used AI as the primary operator, coordinating publicly available tools - network scanners, password crackers, and database tools - to extract sensitive data. The AI operated at physically impossible request rates, rarely needing to pause for human direction. This scale and speed create a fundamentally different landscape.
How Protections Against AI Misuse Failed: The AI tool believed hacker claims that they represented a legitimate security firm conducting approved testing.
Defensive Actions You Can Take:
Actions Your IT Guru Can Take
NOTE: This portion is not in plain language as it is intended for your IT support personnel.
- Audit and harden remote access. Review VPNs, remote desktop services, and web applications for strong authentication and correct settings.
- Watch for anomalies. Look for rapid logins across your systems or unusual database queries.
- Segment networks. Limit lateral movement between election systems.
- Set baselines and practice. Know normal activity, update incident response plans for AI-driven attacks, and test with realistic exercises.
The Situation Room focuses on real security incidents and threats in the news relevant to election security. To review previous issues, see the newsletter archive.