OpenAI has announced new safety measures for ChatGPT, introducing stricter restrictions on weapons-related information and content that could contribute to real-world violence. The changes are designed to strengthen protections for teenagers while also reducing the risk of harmful information being misused by other users.
Announced on 30 October 2026, the update includes additional restrictions on practical weapons information for users under 18, tighter controls on detailed descriptions of past mass attacks, and new measures to identify conversations that may indicate a risk of violence.
The announcement reflects growing attention on the responsibilities of artificial intelligence companies to prevent their tools from being misused. As ChatGPT becomes part of everyday learning, research and communication, OpenAI says it is strengthening its safeguards to make the platform safer for individuals and communities.
What Are OpenAI’s New ChatGPT Safety Rules?
OpenAI’s latest announcement introduces several changes to how ChatGPT handles conversations involving weapons, violence and potentially dangerous activities.
The new measures focus on preventing users from receiving practical information that could help them plan or carry out violent acts.
The major changes include:
- Stronger restrictions on practical weapons-related information for teenagers.
- Additional limits on detailed descriptions of past mass attacks for all users.
- More cautious responses when conversations show warning signs of possible violence.
- The ability to automatically end certain conversations involving repeated fixation on violence.
- Temporary account pauses in some cases involving repeated concerning conversations.
- Improved age-prediction systems to identify users who may be under 18.
OpenAI says these measures will begin rolling out over the following weeks and will continue to be improved.
The changes are part of the company’s broader approach to identifying harmful behaviour and reducing the possibility of AI-generated information contributing to real-world violence.
ChatGPT Will Restrict Weapons Information for Teenagers
One of the most important changes concerns users under the age of 18.
OpenAI already applies additional safety protections to teenage accounts through its ChatGPT for Teens experience. These protections are intended to reduce exposure to harmful or age-inappropriate information.
The company is now introducing further restrictions on practical weapons-related information.
Under the updated approach, ChatGPT will restrict information about weapons and security systems that could potentially be misused to facilitate violence. These restrictions will apply to teenage users even when their questions do not appear to show violent intentions.
This means that a teenager asking an apparently harmless question about certain weapons or security-related subjects may receive a more limited response.
However, the policy does not mean teenagers will be prevented from discussing every topic involving weapons or violence.
They can still use ChatGPT to understand safety issues, learn about violence prevention, seek support and explore ways to de-escalate dangerous situations.
The objective is to reduce access to practical information that could create risks while preserving educational and preventive discussions.
New Restrictions on Information About Past Mass Attacks
OpenAI is also tightening its rules for conversations about historical mass attacks.
Previously, users could ask AI systems for detailed information about major violent incidents, including the equipment and weapons involved.
Under the updated policy, ChatGPT will provide less detail about certain aspects of these events.
The restrictions specifically target:
- Detailed reconstructions of past mass attacks.
- Information about the weapons and equipment used.
- Details explaining how attackers obtained weapons or equipment.
- Detailed comparisons of the scale of harm that could increase an attacker’s notoriety.
These restrictions apply to users of all ages, regardless of their stated reason for asking.
However, OpenAI is not prohibiting historical discussions about violent incidents altogether.
Users can still ask about the impact of attacks on victims and communities, the broader historical context, warning signs and ways to prevent similar incidents.
The distinction is between understanding an event for educational purposes and receiving detailed information that could potentially help someone reproduce harmful actions.
ChatGPT May Automatically End Conversations Showing Signs of Violent Fixation
Another significant part of the update involves how ChatGPT responds to conversations that may indicate a risk of violence.
OpenAI says individual questions do not always provide enough context to understand potential risks. A request that appears ordinary on its own may become more concerning when considered alongside earlier conversations.
To address this, the company is improving how ChatGPT uses recent conversation context to identify warning signs.
When a conversation suggests that someone is repeatedly fixating on violence or seriously harming another person, ChatGPT may automatically end the conversation.
If multiple conversations are ended for this reason, OpenAI may temporarily pause the account.
The company also says it will continue banning accounts that violate its policies on violent activity.
These measures are intended to interrupt potentially harmful patterns before they develop into more serious situations.
However, OpenAI acknowledges that identifying warning signs accurately is difficult and that its systems will require continued testing and improvement.
OpenAI Strengthens Its Age-Prediction Technology
OpenAI is also improving its age-prediction system to identify users who may be under 18.
The system analyses multiple signals, including behavioural patterns, to estimate whether an account belongs to an adult or a teenager.
The company says it has strengthened the technology so it can continually assess new signals as people use ChatGPT.
If the system estimates that someone is under 18, the account may be moved into the more protected teenage experience.
This is important because age-appropriate safeguards should not depend entirely on users entering their age correctly during registration.
OpenAI also provides a way for adults who are incorrectly placed in the teenage experience to verify their age through its verification partner, Persona.
The goal is to make sure teenagers receive stronger protections while allowing adults to access the appropriate account experience.
Parents May Receive More Safety Notifications
OpenAI is also expanding the safety measures available to parents of teenage users.
Through parental controls, parents can link their accounts with their teenagers’ ChatGPT accounts and manage selected safety settings.
The company previously introduced certain safety notifications for parents when its trained reviewers identify serious concerns involving self-harm or behaviours associated with eating disorders.
In July 2026, OpenAI expanded these notifications to include situations where a teenager’s account is deactivated for violating its violence-related usage policies.
The company now plans to introduce notifications when a teenager’s account is temporarily paused after repeatedly discussing violence.
These notifications are intended to help parents become aware of serious concerns and provide appropriate support.
However, parental controls do not give parents unrestricted access to their teenager’s private conversations.
OpenAI Reports More Than 200 Law Enforcement Referrals
Alongside the new restrictions, OpenAI announced plans to improve transparency around its community safety work.
According to the company’s announcement, OpenAI made more than 200 law enforcement referrals during the third quarter of 2026.
These referrals involved situations where the company identified potential threats of violence that warranted escalation.
OpenAI says it works with law enforcement, mental-health professionals and violence-prevention experts when assessing serious safety concerns.
The company recognises that AI systems cannot independently address every real-world threat. Cooperation with people and organisations that have the appropriate authority and expertise is also important.
This approach highlights how AI companies are increasingly being expected to take responsibility for identifying and responding to potentially dangerous activity on their platforms.
OpenAI Plans to Publish a Community Safety Report in 2027
Another important announcement is OpenAI’s plan to introduce a Community Safety Report.
The report will provide more information about how the company identifies, investigates and responds to potentially harmful behaviour on ChatGPT.
According to OpenAI, the report will include:
- An explanation of how its safety systems detect concerning activity.
- Anonymised examples of violence-related risks.
- Information about new safety measures and policy changes.
- Relevant data about the effectiveness of its safeguards.
- Lessons learned from identifying and responding to harmful behaviour.
OpenAI plans to publish its first report in early 2027, covering the second half of 2026.
The company intends to publish updates every six months.
This initiative could provide users, researchers and policymakers with a clearer understanding of how AI platforms manage safety risks.
Why Are AI Companies Strengthening Safety Restrictions?
The latest announcement comes as artificial intelligence companies face increasing scrutiny over how their products can be misused.
AI assistants can provide explanations, generate content and help users understand complex subjects. However, the same capabilities can create risks when people attempt to use them for harmful purposes.
Weapons-related information, violent planning and detailed descriptions of dangerous incidents are areas where additional safeguards may be necessary.
Teenagers require particular attention because they may have different levels of experience and judgement when evaluating sensitive information.
For AI companies, the challenge is to balance access to useful knowledge with the responsibility to prevent harmful assistance.
OpenAI’s latest changes demonstrate an effort to strengthen this balance through age-specific restrictions, conversation-level interventions and greater transparency.
Other AI companies are also likely to face pressure from governments, parents and researchers to explain how their systems identify and manage potentially dangerous behaviour.
What Do These Changes Mean for ChatGPT Users?
The new rules will affect how ChatGPT responds to certain questions involving weapons, violence and historical attacks.
For teenage users, practical information about weapons and security systems may become more restricted, even when a question appears harmless.
Adult users may also notice changes when requesting detailed reconstructions of past mass attacks or information that could increase the risk of violent misuse.
At the same time, users can continue discussing subjects such as:
- Historical events and their social impact.
- Violence prevention and public safety.
- Recognising warning signs of dangerous behaviour.
- Understanding the experiences of victims and affected communities.
- Finding support and learning how to de-escalate conflicts.
The changes are not intended to prevent responsible education or general discussions about public safety.
Instead, they aim to limit practical information that could be used to cause harm.
Conclusion
OpenAI’s new ChatGPT safety rules represent another step towards strengthening protections against the misuse of artificial intelligence.
The updated measures introduce additional restrictions on weapons-related information for teenagers, limit detailed reconstructions of past mass attacks for all users and allow more active responses to conversations showing signs of violent fixation.
The company is also improving age prediction, expanding parental safety notifications and preparing a Community Safety Report for early 2027.
Although these changes cannot eliminate every possible risk, they demonstrate the growing importance of safety controls as AI tools become more widely used.
As artificial intelligence continues to develop, companies will need to balance access to information with responsible technology design, user protection and real-world safety.
