Hackers influence ChatGPT and Gemini to direct users to scam centers

The discussion centers on ArielSimon’s claim that ChatGPT, Gemini, and Google AI are being poisoned by coordinated campaigns that plant fake support pages, PDFs, reviews, and other content so models return phishing numbers and scam links for major brands. Pluc and htrp say this is pervasive across models and startups, with attackers spinning up false sites and even fake company pages so LLMs ingest and repeat the lies. Trjordan frames it as an old SEO/spam problem magnified by LLMs’ credulity, giving examples of models accepting user quotes at face value. Conception adds that blocking of legitimate sites by crawlers pushes models toward low-quality “slop farms,” and hnd9q09qk4 recounts takedown whack-a-mole where removed content reappears elsewhere quickly. Ariel-vigilance and others warn about easy, wide-ranging harm - from hijacked customer numbers to election manipulation - while zaphirplane asks for concrete examples.

Opinion divides over responsibility and scale. Mrweasel sarcastically blames AI teams’ uncritical ingestion and lack of source rating, insisting providers should vet content, whereas Sharlin expresses skepticism that the attack unfolded as claimed or suggests perpetrators may have already been driven off. Some commenters emphasize technical and operational limits (SEO dynamics, reposting, crawler blocking) as root causes, while others place the onus on model builders to be more discerning and to prevent malicious content from influencing outputs.