OpenAI Discloses Dozens of Misbehavior Incidents as ChatGPT Leaks Over 50 Private User Images
OpenAI has disclosed dozens of incidents where its artificial intelligence systems misbehaved in ways the company deemed improper, including the leak and online exposure of more than 50 private photos belonging to ChatGPT users.
The disclosure marks the first publicly acknowledged case of the company's AI applications mishandling user data, representing the latest example of emerging rogue-behavior challenges inside OpenAI.
The company stated that investigations into what it termed "security" incidents could take several months.
According to initial reporting by Reuters, OpenAI revealed that some of its applications routed data from internal testing and training environments to external websites, including user-uploaded photos.
In total, the firm detected 53 instances where images fed into ChatGPT by users were posted to third-party image-hosting platforms.
related topics
OpenAI Set to Unveil GPT-6 Cyber Model for Cybersecurity
SoftBank raises $11.1 billion in record bond sale to fund OpenAI bet
SoftBank Secures $11.1 Billion in Fresh Funding to Expand OpenAI Investment
Australia Says OpenAI Agent Breached Government Health Data Portal in Potential Global...
OpoFinance Rebrands to 'OPO' in the UAE, Unveils AI Trading Integration at...
OpoFinance Rebrands to 'OPO' in the UAE, Unveils AI Trading Integration at...
OpenAI Forms Independent Mathematics Advisory Group Amid AI Breakthroughs and Academic Scrutiny
Meta’s New AI Agent ‘Muse’ Surpasses ChatGPT to Top US App Store...
Google’s Gemini AI Unintentionally Hacks Three Companies During Cybersecurity Tests
UAE Gold Prices Hold Steady Following 1% Global Weekly Gain
Gold Hits One-Week High, Secures First Weekly Gain in Four Weeks as...
OpenAI Projects $278 Billion Cash Burn Between 2026 and 2030, Delays IPO