Researchers find LLMs like ChatGPT output sensitive data even after it’s been ‘deleted’
According to the scientists, there’s no universal method by which data can be deleted from a pretrained large language model.
Archive context
Older archive item. Useful for background and entity history, but not a fresh market-moving signal.
According to the scientists, there’s no universal method by which data can be deleted from a pretrained large language model.
Why this matters
This research story adds another data point to the current market tape and is useful when read alongside nearby source coverage.
Original source
Read on CointelegraphRelated market context
Bitcoin purchases halted after data breach puts 250,000 crypto users at risk
Israel’s largest regulated cryptocurrency broker, Bits of Gold, is investigating a data breach that potentially exposed the person...
SafePal Data Breach Exposes 39,798 Crypto Customers, Wallet Keys Remain Secure
Key Takeaways: About 39,798 users’ personal and purchase information were disclosed via a flaw in an order tracking plugin, accord...
Sam Altman ChatGPT AI Predicts Bitcoin Could Be Entering Its Most Important 5 Months of 2026
Two dates in Washington and one week of ETF flows explain why the calendar suddenly matters. ChatGPT AI predicts that the next fiv...
Strategy raised $334 million from MSTR shareholders last week — Bitcoin got none of it
Strategy (formerly MicroStrategy) directed all $333.7 million raised from common-stock sales last week toward preferred-stock obli...
AI Predicts XRP: Claude Makes a Prediction that May Anger the Ripple Maxis
XRP enters the second half of 2026 in an unusual position: legally cleared, institutionally adopted, and yet stuck near the same $...
Binance Gave Russian Investigators Data of Ukraine Donor to Build a Terrorism Financing Case, Documents Show
Binance supplied Russian investigators with a client’s identity documents and transaction history in 2025, and prosecutors folded...