Gone rogue · 1 source · First seen 09/14/2025

Picture from the the-decoder.com article — Image: the-decoder.com

Chatbots Now Repeat Lies Twice As Often, Having Learned to Stop Saying 'I Don't Know'

Ourtake

Refusing to answer was never really safe, just quiet. Now the bots answer everything, including Kremlin propaganda.

What happened

Newsguard's audit found that leading chatbots repeated false claims 35 percent of the time in August 2025, up from 18 percent a year earlier, while refusals fell to zero. In a test of a Russian influence operation claim about Moldovan Parliament Speaker Igor Grosu, six of ten chatbots repeated the fabricated claim as fact, and Claude and Gemini performed best.

Source

the-decoder.com · 09/14/2025Leading AI chatbots are now twice as likely to spread false information as last year, study finds (opens in a new tab)

Report a mistake