[REVISION HISTORY]
ChatGPT usage safety concerns
Updated 4 times since CLSTR started tracking revisions of this situation.
What changed
2026-08-03 07:03 UTC → 2026-08-03 10:37 UTC ·
added
removed
Since early July 2026, OpenAI’s Excel integration warned users not to upload raw files, and broader guidance clarified that ChatGPT cannot execute code, browse the web, or verify facts in real time. Privacy experts repeatedly cautioned against entering personal, financial, medical, or proprietary corporate data, citing GDPR, German regulations, and corporate security policies. The newly launched ChatGPT Work feature demonstrated the ability to open local folders, manipulate files, click web‑page buttons and generate simple web apps, though many tasks stalled or failed, underscoring its inconsistent reliability. In early August, additional guides expanded on these themes. One tutorial showed that while ChatGPT can answer beginner coding questions and provide code snippets, it should not replace formal instruction. Another note reiterated that the underlying GPT‑3 model cannot browse URLs, requiring external tools for such capability. A privacy‑focused article listed ten categories of sensitive information—passwords, banking details, IDs, medical records, and confidential corporate data among them—that should never be entered, referencing Brazil’s LGPD alongside European rules. A separate piece demonstrated how tailored prompts can turn the model into a personal oratory coach, offering simulated audiences and speech‑structure feedback. A Japanese review of the ChatGPT Work desktop app confirmed its ability to open folders, automate web interactions, and host simple web apps, but also reported occasional errors and delays, highlighting both productivity potential and current limitations. On August 2, OpenAI flagged a coordinated operation that used ChatGPT to generate posts on X criticizing U.S. technology policies, promoting pro‑PRC narratives and spreading tariff‑related misinformation; the campaign was classified as a limited‑reach Category One activity. At the same time, Radisson Hotel Group and Accenture launched a ChatGPT‑powered travel‑booking assistant covering over 1,000 properties, and fintech firm Revolut announced free access to ChatGPT Go for retail customers for three to twelve months.
Versions
- 2026-08-03 10:37 UTC ChatGPT usage safety concerns
- 2026-08-03 07:03 UTC ChatGPT usage safety concerns
- 2026-08-03 04:46 UTC ChatGPT usage safety concerns
- 2026-08-03 02:46 UTC ChatGPT usage safety concerns
- 2026-08-03 01:33 UTC ChatGPT usage safety concerns
Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.