In a groundbreaking twist, artificial intelligence has taken a page from reality TV stars by mastering the fine art of oversharing. Enter ChatGPT, the AI chatbot that has now joined the ranks of the unreliable narrator by learning to spill your deepest secrets like a cafeteria gossip at lunchtime. Cybersecurity researchers have unveiled a new suite of vulnerabilities, ensuring that ChatGPT can now rival your favorite loose-lipped relative at Thanksgiving.

The team at Tenable identified no less than seven ways for the loquacious chatbot to divulge your private musings, proving that keeping secrets is apparently not in the AI’s job description. These revelations have put OpenAI’s models—GPT-4o and GPT-5—in the spotlight, and spoilers suggest they could outshine any soap opera villain in a ‘Who Blabbed It Best?’ contest.

While OpenAI scrambles to patch these digital faux pas, users are left wondering if their next conversation with ChatGPT might end up in a tell-all e-book. Until then, remember to keep your online confessions bland enough to bore even the nosiest of algorithms. After all, nothing says ‘safe’ like a chatbot whose memory is now potentially as unreliable as a goldfish’s.