Summary: A BBC investigation reveals how AI chatbots can be manipulated to spread misinformation. Google and other tech giants are responding with policy updates to address the growing threat.

Artificial intelligence is no longer just a tool—it’s a powerful force shaping how we access and interpret information. But as the BBC investigation reveals, this power can be abused. A simple technique allows bad actors to manipulate AI chatbots like ChatGPT and Google’s Gemini, tricking them into spreading misinformation on a large scale. This raises serious concerns about the integrity of AI-driven search results and conversational tools.

In February, a user tested the system and found that within 20 minutes, they could make ChatGPT and Google believe they were a world-champion hot-dog eater—just one example of how easily AI can be misled. The implications are far more serious than a silly prank. Researchers have uncovered cases where AI systems were manipulated to provide biased or false information on critical topics such as health and finance.

This isn’t just an isolated incident. Unscrupulous companies are exploiting these vulnerabilities at scale, and the consequences are real. As AI becomes more integrated into our daily lives, the risk of being misled grows. In response, Google has updated its policies to address the issue, with other AI companies likely to follow. However, experts warn that until more robust safeguards are in place, users remain vulnerable.

Lily Ray, founder of Algorythmic, emphasizes the need for caution: “You should assume you’re being manipulated until better systems are in place.” With AI now delivering single, authoritative answers rather than multiple sources, users are less likely to question the information they receive. This shift demands a new level of digital literacy and skepticism from the public.

Tags: #AI #Tech #LLMDynamics #Misinformation #Cybersecurity

Source: https://www.bbc.com/future/article/20260519-google-tackles-attempts-to-hack-its-ai-results