OpenAI repeatedly re‑enables ‘allow training’ setting, says HN
openai training
| Source: HN | Original article
OpenAI repeatedly reactivates the “allow training” option, overriding users’ opt‑out settings across their accounts.
OpenAI’s data‑control toggle for model training has resurfaced as a point of contention on Hacker News, where users report that the company repeatedly re‑enables the “allow training” option after it has been turned off. The complaints echo earlier concerns about the opacity of OpenAI’s data‑use policies and the difficulty of maintaining a consistent opt‑out stance.
OpenAI’s Help Center states that disabling model training “applies to your entire account” and that the setting is respected across devices. However, the same documentation notes that if a user later switches the toggle back on, “future conversations resume being eligible for training.” A separate “Temporary Chat” mode guarantees deletion after 30 days and is never used for training, even when the main toggle is active. For personal accounts on ChatGPT Free, Plus or Pro plans, data sharing is enabled by default but can be opted out of, while enterprise users are directed to a dedicated privacy page.
The issue matters because the ability to permanently exclude data from future model improvements is a cornerstone of user trust, especially for researchers and businesses handling sensitive information. Prior reporting highlighted similar doubts about OpenAI’s handling of unpublished academic work and the broader reliability of its training data practices. If past contributions remain in training sets after an opt‑out—something the company’s own disclaimer acknowledges—users may find their data inadvertently influencing new models despite their preferences.
What to watch next is OpenAI’s response to the renewed criticism. Stakeholders will be looking for clearer, enforceable mechanisms that prevent inadvertent re‑enrollment, as well as any regulatory scrutiny that could arise from perceived inconsistencies. The community will also monitor whether the company updates its Help Center or introduces new safeguards for both personal and enterprise users.
Sources
Back to AIPULSEN