OpenAI launches a ChatGPT Powerpoint plugin and warns it might accidentally delete your content

The buried detail here is not the feature itself but the admission: OpenAI is shipping a tool into enterprise workflows while explicitly acknowledging it may destroy user content, with no public disclosure of how frequently that occurs or under what conditions.

Modelwire has no prior coverage to anchor this to directly, so this sits largely disconnected from recent activity in our archive. The story belongs to a broader pattern of AI assistants being pushed into document-editing contexts before reliability baselines are established. The honest framing is that this is an early-access integration where the vendor has offloaded risk disclosure onto the user rather than solving the underlying problem. That pattern matters because enterprise buyers evaluating Microsoft 365 Copilot competitors will now have to weigh whether a ChatGPT plugin with acknowledged destructive failure modes belongs in the same stack as tools handling irreplaceable business documents.

Watch whether Microsoft or OpenAI publish a concrete error-rate figure or a rollback mechanism before the plugin exits early access. If neither appears by general availability, that signals the data-loss risk is being treated as acceptable rather than solvable.