
The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Researchers at Zafran Labs disclosed four critical vulnerabilities, dubbed "DifyTap," in the open-source AI platform Dify. These flaws allow unauthenticated attackers to access and exfiltrate private data, AI conversations, and documents across tenants, impacting over a million AI applications used by major companies worldwide.[AI generated]
Why's our monitor labelling this an incident or hazard?
The event involves an AI system (Dify, an AI orchestration platform) whose vulnerabilities could lead to unauthorized access and leakage of sensitive AI chat data and documents. This constitutes a violation of privacy and potentially human rights, which fits the definition of harm under AI Incident (c). The vulnerabilities have been exploited in theory, and the harm is direct if an attacker uses these flaws. Therefore, this qualifies as an AI Incident rather than a hazard or complementary information, since the vulnerabilities directly enable harm through the AI system's malfunction or security flaws.[AI generated]