Meta has fixed a security rapid of the meta Ai chatboot users to receive access and the private prompts and ai-generated answers of other users.
Sand Herice Hodkasia, the safety of the safety protr, excellence, excessary, the metal, has not paid for the 24,000 up-to-date, to fix, fasten, 20.24.
Meta has taken the 254. January 2525 dedicated, and has Hodkasia, and found no proof that the mistake was exploited.
Hodkasia tells techniquerunt that he identified the mistake after examining them as Meta ai Allow its logged in users to change their AI prompt to re-generate text and pictures. He discovered that when a user return their prompts, the meta-end server server is the prompt and her ai-generated response to get a unique number. Analyze the network traffic and his browser at the editing, Hodkasia found that he has the unique number and the meta’s servers and ai-generated response from another.
The bug made that the meta is not properly checked right, to ensure that the user is the prompt and its answer is authorized to see it. Hodkasia told the prompt numbers of the meta’s server
If it has on the appscunnulon, the meta confirmed the mistakes “the proof of the proof of abuse of abuse,” metachers, “Metare-speaker has signed the mete-speaker.
News of the error of the error comes at a time when tech gates scrambling are scrambling and to catch their ai products in spite of Many security and privacy risks associated with their use.
Meta ai’s standalone app, what Debuting previously this year To compete with rival apps like chatgt, launched on a rock of rock after some users Inadvertently publicly shared what they thought private conversations by chatbot.
