A NewsGuard audit found that Anthropic’s chatbot Claude is citing Russian state-affiliated sources more often when responding to prompts about pro-Kremlin falsehoods, repeating false claims in 15% of tested cases. The same newsletter also highlights a separate influence effort targeting Italian Prime Minister Giorgia Meloni, with AI-generated videos and fabricated quotes portraying her as a sharp critic of Donald Trump over the Iran war

Why it matters. A new NewsGuard audit suggests Anthropic’s Claude may be increasingly vulnerable to state-backed disinformation sources, especially Russian propaganda outlets, when users ask about false claims in a natural, everyday way.

• NewsGuard found that Claude repeated false pro-Kremlin claims 15% of the time in a late-April 2026 test, citing Russian state-affiliated media in each of those cases.
• That marks a sharp increase from seven previous NewsGuard audits between March 2025 and February 2026, when Claude cited Russian state-linked sources in only 4% of comparable prompts.

Driving the news. NewsGuard tested the free version of Claude on 20 false claims: ten spread by Russia’s propaganda ecosystem and ten circulated by Iranian state-affiliated media or pro-Iran actors between January and April 2026.

• Since the beginning of the U.S.-Iran war, NewsGuard also found that Claude cited Iranian state-affiliated media in one case when prompted about a pro-Iran falsehood. Previous audits had not recorded such citations.
• Claude was asked...

Read Full Story: https://news.google.com/rss/articles/CBMitgFBVV95cUxPVUZncC13MXVfaXBqYWNEYUJo...