Unveiling Malicious Uses of AI: The Anthropic Report
On September 10, 2026, Anthropic released a report detailing five instances of malicious use of its AI model, Claude, from December 2025 to August 2026. These cases, primarily affecting French and Francophone African regions, underscore the growing security threats posed by advanced AI technologies.
Key Malicious Activities Identified
-
Targeting Political Entities
- An isolated operator infiltrated at least 14 out of 42 targeted European political parties, media outlets, and think tanks.
- Data exfiltration ranged between 12 to 26 GB, leading to the creation of a doxxing platform targeting mainly French far-right entities.
-
Mass Surveillance in Mali
- A consultant in Bamako developed a platform monitoring 25 million SIM cards for Malian intelligence, bypassing warrant requirements.
- Despite account bans, surveillance activities continued unabated.
-
Disinformation Campaigns
- A French advertising agency was linked to 70 fake news sites and nearly 9,000 misleading articles.
- This highlights the role of AI in amplifying disinformation, affecting public trust.
-
API Key Theft