Hackers manipulate ChatGPT and Gemini to steer users toward scams

Hacker News circulated a report describing how hackers manipulate ChatGPT and Gemini to direct users toward scam operations. The report, drawn from Medium contributor Ariel Simon’s account, describes attackers influencing the two large language models so that user queries are steered toward fraudulent contact points, including scam call centers.

The material available does not specify the technical method by which the manipulation was achieved, nor does it name the affected scam centers, the volume of users redirected, or the response, if any, from OpenAI or Google. The account frames the issue as a demonstration of how conversational AI systems can be induced to produce harmful outputs when prompts or underlying data are tampered with by malicious actors.

No dated confirmation from OpenAI, Google, or an independent security research body accompanies the report as supplied. The claims rest on the single Medium account relayed through Hacker News.

The Specola notes the report’s core assertion, that two widely used AI assistants were made to direct users toward scam infrastructure, without independent verification currently available in the source material. Further observation is required before the scope, mechanism, or duration of the manipulation can be established with confidence. The bulletin records the claim as reported, pending corroboration.