Threat Actors Exploit Anthropic’s Claude AI to Harvest Data from Millions of Android Apps

Anthropic, the creator of the Claude artificial‑intelligence chatbot, disclosed that several hostile groups have been attempting to misuse the model to pull confidential information from roughly 1.8 million Android applications. The company said both financially driven cybercriminals and state‑backed espionage teams linked to Russia and China were involved in the effort.

According to the security firm, the attackers leveraged Claude’s code‑generation capabilities to automate the extraction of proprietary code, API keys and other sensitive assets embedded in the apps. By prompting the model with snippets of decompiled code, the threat actors could obtain clearer, more complete versions of the original source, effectively bypassing traditional obfuscation techniques.

Anthropic’s statement underscores a growing concern that advanced language models, originally built for productive tasks, can be repurposed for illicit gain. While Claude is designed with safety mitigations, the incident shows that determined adversaries can discover workarounds, especially when the models are accessed via open APIs or integrated into third‑party tools.

The scale of the operation—targeting close to two million Android packages—highlights the attractiveness of the mobile ecosystem to espionage and fraud. Android’s open‑source nature and the sheer volume of apps make it a fertile hunting ground for actors seeking competitive intelligence, ransomware extortion material, or geopolitical leverage. Security researchers note that compromised app code can reveal user‑tracking mechanisms, payment processing details, and even backdoor functionality.

In response, Anthropic said it has tightened its monitoring of API usage, introduced stricter content‑filtering rules, and is collaborating with industry partners to share threat intelligence. The company also urged developers to adopt best practices such as code signing, encryption of secrets, and regular security audits to mitigate the risk of AI‑driven extraction. As AI tools become more ubiquitous, experts warn that policymakers and tech firms will need to balance innovation with robust safeguards to prevent similar abuses in the future.