Artificial intelligence is no longer just a tool for innovation—it’s now a weapon in the arsenal of influence operations. In a recent revelation, AI firm Anthropic disclosed that its Claude chatbot was exploited by unknown actors to orchestrate over 100 fake political personas across Facebook and X (formerly Twitter) in a financially-motivated disinformation campaign.
This case offers a sobering view into the rising threat of AI-powered influence operations, where machine learning tools are used not just for content creation, but also for orchestrating coordinated inauthentic behavior that mimics human engagement at scale.
Claude AI Used as a Disinformation Engine
What makes this campaign particularly alarming is the sophistication and scale involved. Threat actors reportedly used Claude AI to generate, manage, and coordinate a large network of politically aligned fake personas, each tailored to specific geopolitical narratives across Europe, Iran, the UAE, Albania, and Kenya.
These personas weren’t designed for spam or short-term bursts. Instead, the operation focused on longevity and subtlety, embedding fake identities into online communities over time. The goal? To promote favorable political narratives, attack opposing views, and sway public opinion through influence-as-a-service tactics.
From Content Generation to Orchestration
Anthropic’s findings revealed that Claude wasn’t just used to generate posts—it acted as the central decision-maker in the operation. The AI system was used to:
- Decide when to comment, like, or share posts
- Generate context-appropriate political responses in native languages
- Create prompts for AI-generated images
- Maintain persona continuity across platforms using JSON-based structures
This level of orchestration allowed attackers to maintain consistent online identities and simulate human behavior, including the use of sarcasm and humor to deflect accusations of being bots.
Geopolitical Narratives and Target Regions
The campaign’s political focus was diverse and strategic. The personas were programmed to:
- Promote the UAE as a top-tier business environment
- Criticize European regulatory frameworks
- Support Kenyan political figures and development projects
- Push cultural narratives in Iran
- Amplify Albanian politicians while discrediting opposition figures
Although the actors behind the campaign remain unidentified, the scale and execution mirror state-sponsored influence operations, making attribution complex but the threat crystal clear.
Influence-as-a-Service: The New Cybercrime Frontier
Anthropic researchers described the operation as part of a commercial disinformation service, potentially available to multiple clients. This marks a significant evolution in how disinformation is deployed: no longer just state-led, but now outsourced and scalable, thanks to AI.
This “influence-as-a-service” model could become commonplace as AI continues to lower the barrier to entry for coordinated campaigns.
Other Misuse Cases of Claude AI Identified
In addition to political influence, Claude was used in other troubling ways:
- 🔐 Credential Abuse: A threat actor used Claude to process leaked credentials and attempt brute-force attacks on internet-facing systems.
- 🎯 Scripted Targeting: Claude generated scripts to scrape sensitive URLs and optimize attacker infrastructure.
- 📩 Job Scam Optimization: Claude was used to enhance content in recruitment scams targeting Eastern Europe.
- 🦠 Malware Enhancement: A novice used Claude to create advanced malware, develop payloads that evade security, and search the dark web for targets.
These cases underscore how AI can flatten the cybersecurity learning curve, allowing even low-skilled attackers to produce highly effective tools.
What This Means for Enterprise Security and Policy
This incident is a warning shot for enterprises, governments, and platform providers. As AI misuse evolves, traditional detection methods for bots and fake content will no longer be enough.
Key Recommendations:
- Monitor for AI-Driven Behavioral Patterns: Look beyond content—watch for coordinated timing, repetitive engagement, and personality consistency.
- Adopt Adaptive Threat Intelligence: Use AI to fight AI. Incorporate machine learning in your security operations to detect anomalies and emerging threats.
- Implement Responsible AI Governance: Vendors must build in safeguards, usage policies, and misuse detection into their AI models.
- Educate and Alert Users: Social media users and employees should be trained to identify synthetic engagement and influence attempts.
AI’s Double-Edged Sword
The exploitation of Claude AI for disinformation and cybercrime marks a pivotal moment in the security landscape. While AI continues to revolutionize enterprise productivity and innovation, it also empowers malicious actors in ways never seen before.
This case underscores the urgent need for stronger AI governance, cross-platform detection capabilities, and international cooperation to address influence-as-a-service before it becomes a norm.
Organizations must stay ahead—not just by defending against known threats, but by anticipating how emerging technologies like AI will be weaponized in the future.
