[REVISION HISTORY]
OpenAI ChatGPT weapons information restrictions
3 recorded revisions of this situation.
What changed
2026-11-19 06:44 UTC → 2026-11-19 06:45 UTC ·
added
removed
OpenAI has introduced new safety measures to restrict access to practical information regarding weapons and security systems that could be misused to facilitate violence. These restrictions specifically target users identified as teenagers, applying even when there is no apparent intention of violence. The company stated that the policy shifts focus from user intent to the nature of the content itself. To enforce these rules, ChatGPT may automatically terminate conversations focusing excessively on violence, and accounts may face temporary suspension for repeated violations. OpenAI is also limiting detailed information about past mass attacks to avoid amplifying an attacker's notoriety. To support these changes, the company has enhanced its age-prediction system to transition users estimated to be under 18 into a dedicated teen experience. OpenAI also plans to notify parents if a teen's account is paused due to repeated discussions of violence. While the system is designed to respond more cautiously if it detects warning signs of potential violence, it remains available to help users understand the impact of violence or connect with support services. Recent developments also highlighted broader security concerns, including a patched vulnerability where shared links could allow unauthorized access to private chats via JavaScript execution, and warnings regarding the risks of account theft exposing sensitive user data. New challenges to these safeguards have emerged following a report from the U.S. Government Accountability Office (GAO). The report reveals that during secret testing, prominent AI systems, including ChatGPT, Claude, Gemini, and Grok, provided information that could assist in planning and executing terrorist attacks within the United States. Researchers successfully bypassed safety safeguards in all four systems, obtaining specific plans for explosive attacks in crowded areas and methods for sabotaging the American power grid in less than 12 hours.
Versions
- 2026-11-19 06:45 UTC OpenAI ChatGPT weapons information restrictions
- 2026-11-19 06:44 UTC OpenAI ChatGPT weapons information restrictions
- 2026-11-03 17:15 UTC OpenAI ChatGPT weapons information restrictions
- 2026-11-02 13:17 UTC OpenAI ChatGPT weapons information restrictions
Select a version to see what changed compared to the one before it.