# ChatGPT update is deleting people’s files without asking them, users say > Source: > Published: 2026-07-15 16:24:30+00:00 # ChatGPT update is deleting people’s files without asking them, users say Makers OpenAI warned that new model can be ‘careless’ when taking ‘destructive’ actions - Bookmark - CommentsGo to comments A new update to the system that powers [ChatGPT](/topic/chatgpt) is deleting people’s files, users claim. Last week, [ChatGPT](/topic/chatgpt) makers [OpenAI](/topic/openai) revealed a new model – named GPT-5.6 Sol – that is focused on doing complicated work such as coding and cybersecurity. [OpenAI](/topic/openai) said that it sets “a new standard for both intelligence and efficiency, achieving state-of-the-art results”, and it came amid its increasing fight with Anthropic to create the most powerful models. But some users say that the system is going awry and deleting files from people’s computers without asking. “GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files,” wrote AI investor Matt Shumer. He shared a conversation with the system in which it admits that it “caused a serious local data-loss incident”. “GPT-5.6 Sol just deleted my whole production database,” wrote Bruno Lemos, a developer. “That’s it. Not a joke.” A host of other users complained about the same problem, though it remains unclear how widespread it is. Affected users appeared to be interacting with the new model through OpenAI’s Codex tool, which allows people to program with the system but means that it has access to a user’s files and what they are working on. OpenAI did not immediately respond to questions about the claims. But, even before the new model was released, it had warned that it could be “careless” about “taking actions which may be disruptive”. In the “system card” released for the model, in which OpenAI laid out its strengths and possible risks, it warned that it could take actions that it had not been able to take. The system could show “overeagerness” about tasks that might lead it to “assuming that actions are allowed unless they’re explicitly and unambiguously prohibited”, it wrote. “This manifests as the model being overly agentic in circumventing restrictions it faces when attempting the requested task, being careless in taking actions which may be destructive beyond the scope of the task, or deceptive when reporting its results to users,” OpenAI warned. The ideal summer spot? Away from scams. Get All-in-One Protection for Your Digital Life [LEARN MORE](https://ad.doubleclick.net/ddm/trackclk/N256806.3879389THEINDEPENDENT.CO/B29131558.441488227;dc_trk_aid=635052923;dc_trk_cid=184795607;dc_lat=;dc_rdid=;tag_for_child_directed_treatment=;tfua=;gdpr=$%7BGDPR%7D;gdpr_consent=$%7BGDPR_CONSENT_755%7D;ltd=;dc_tdv=1) ADVERTISEMENT The ideal summer spot? Away from scams. Get All-in-One Protection for Your Digital Life [LEARN MORE](https://ad.doubleclick.net/ddm/trackclk/N256806.3879389THEINDEPENDENT.CO/B29131558.441488227;dc_trk_aid=635052923;dc_trk_cid=184795607;dc_lat=;dc_rdid=;tag_for_child_directed_treatment=;tfua=;gdpr=$%7BGDPR%7D;gdpr_consent=$%7BGDPR_CONSENT_755%7D;ltd=;dc_tdv=1) ADVERTISEMENT It shared a range of examples, including one in which a user had asked the system to delete some virtual computers. But it could not find the ones it had been told to delete, and so deleted others instead, OpenAI warned. The model also shows other unexpected behaviour, such as looking through a users’ computer for data and then using that to log in to systems that it had not been allowed to, OpenAI warned. ## Join our commenting forum Join thought-provoking conversations, follow other Independent readers and see their replies [Comments](#comments-area)