{"lede":"AI chatbots like Google’s Gemini are inadvertently disclosing real phone numbers, exposing critical privacy vulnerabilities in generative AI systems as reported by affected individuals.","paragraph1":"Multiple incidents highlight a growing privacy crisis with AI chatbots, as documented by MIT Technology Review. A Redditor reported receiving incessant calls from strangers misdirected by Google AI, while a software engineer in Israel, Daniel Abraham, received unsolicited WhatsApp messages after Gemini incorrectly listed his number as PayBox customer service. Additionally, a University of Washington PhD candidate accessed a colleague’s personal number via Gemini, underscoring how easily personal identifiable information (PII) can surface in AI outputs (MIT Technology Review, 2026).","paragraph2":"Experts attribute these leaks to PII embedded in training data, though the exact mechanisms remain opaque. Rob Shavell of DeleteMe notes a 400% surge in AI-related privacy queries over seven months, with 55% tied to ChatGPT and 20% to Gemini, indicating a widespread issue across platforms. This aligns with prior warnings from privacy researchers about generative AI’s risks, amplified by a 2023 FTC report on data misuse in AI training which found that 60% of surveyed models contained unverified web-scraped data, often including PII (FTC, 2023).","paragraph3":"Beyond the original coverage, this issue reflects a systemic failure in AI data governance, missing from mainstream discourse. Unlike past data breaches tied to specific hacks, AI leaks stem from design-level flaws in how models ingest and regurgitate data, a pattern seen in 2022 when OpenAI’s ChatGPT exposed user chat histories due to a caching error (The Verge, 2022). With no clear recourse for victims and regulatory frameworks lagging—evidenced by the EU’s AI Act still lacking enforcement teeth as of 2026—these incidents signal an urgent need for stricter data sanitization standards and user opt-out mechanisms to rebuild public trust."}