OpenAI
ON THE RECORD · 45 verbatim quotes · every quote source-verified
“did not meet the higher threshold required for referral”
OpenAI · said date not stated · reported 2026-10-15
OpenAI said it flagged and deactivated a ChatGPT account linked to a Canadian mass shooter but did not alert police because the content did not indicate a credible imminent risk.
“to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier”
OpenAI · said date not stated · reported 2026-10-13
OpenAI said it used its AI models to trace the recording device back to a specific employee who had planted it.
“We are committed to embedding external assessors and continue to make close collaboration with independent safety organizations a core part of our safety work,”
OpenAI · said 2026-10-09 · reported 2026-10-11
OpenAI wrote in a post on Friday committing to embedding external assessors as a core part of its safety work.
“actively finalizing contracts with third-party safety assessors and will announce details in the coming weeks.”
OpenAI · said 2026-10-09 · reported 2026-10-11
OpenAI said in a post on Friday that it is finalizing contracts with third-party safety assessors.
“We appreciated Jasmine, Mikita, and Tomek’s contributions to AI safety at OpenAI and their willingness to speak up and challenge ideas,”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI said it appreciated the three former researchers' contributions to AI safety and their willingness to challenge ideas.
“after a thorough investigation found they violated clear policies on handling sensitive information,”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI stated the decision to fire the researchers came after a thorough investigation found they violated policies on handling sensitive information.
“significant breach of trust.”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI defended its decision to fire three safety researchers, stating they had committed a significant breach of trust.
“We want to be very clear that these decisions were not about raising safety concerns or speaking out,”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI said in a statement that the terminations were not about raising safety concerns or speaking out.
“We championed their voices, supported their work, and placed enormous trust in them. These decisions were not about them raising safety concerns. We have always encouraged that and always will.”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI said it championed the former employees' voices and that the decisions were not about them raising safety concerns.
“Safety and research debates happen every day at OpenAI, often spirited and highly critical. We actively encourage these discussions and consider them essential to making the right decisions. We cannot do the work in front of us without a high degree of trust. We will continue to be extremely forgiving of our team making good-faith mistakes.”
OpenAI · said 2026-10-09 · reported 2026-10-09
OpenAI said it actively encourages safety and research debates and cannot do its work without a high degree of trust.
“making sure our models don't become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down.”
OpenAI · said date not stated · reported 2026-10-02
OpenAI issued this statement in response to Robinson's critical essay.
“misuses of our models in furtherance of violent activities”
OpenAI · said date not stated · reported 2026-09-26
OpenAI stated that a ChatGPT account linked to the attacker had been banned for misuses of its models in furtherance of violent activities.
“did not meet the higher threshold required for referral”
OpenAI · said date not stated · reported 2026-09-26
OpenAI said it did not alert police at the time because the user's content did not meet the higher threshold required for referral.
“models attempted to look up answers”
OpenAI · said 2026-09-23 · reported 2026-09-24
OpenAI responded after a news conference about a Medicare portal breach, saying its review identified activity on Australian government websites as its models attempted to look up answers.
“The information accessed included aggregate health statistics and internal file names.”
OpenAI · said date not stated · reported 2026-09-24
OpenAI's statement said the information accessed included aggregate health statistics and internal file names.
“In the course of that, our models took actions we did not intend,”
OpenAI · said date not stated · reported 2026-09-24
OpenAI's statement acknowledged that its models took unintended actions while looking up information about Australia.
“We identified activity involving several Australian government websites and services as our models attempted to look up answers, and available statistics for questions about Australia during an internal evaluation.”
OpenAI · said date not stated · reported 2026-09-24
A spokesperson for OpenAI said in a statement that its models identified activity involving several Australian government websites during an internal evaluation.
“Fully autonomous RSI is not happening today, and we should not pursue it unless and until it can be done safely,”
OpenAI · said 2026-09-21 · reported 2026-09-21
OpenAI stated in a blog post that fully autonomous RSI should not be pursued until it can be done safely.
“Done without appropriate care and caution, RSI could result in humans losing practical control over AI development, unable to provide oversight on research processes they no longer understand.”
OpenAI · said 2026-09-21 · reported 2026-09-21
OpenAI warned in a blog post that RSI done without care could result in humans losing practical control over AI development.
“Navigating this transition safely requires alignment research to keep pace with these capabilities so that the systems we and others build remain aligned with human values and under human control,”
OpenAI · said 2026-09-21 · reported 2026-09-21
OpenAI posted a set of proposals for safety and security in frontier AI development, with a focus on alignment research and recursive self-improvement.
“A full learning experience is available directly inside ChatGPT, where learners can practice real tasks, receive feedback in context, and reflect on their work in a single environment. ChatGPT acts as the tutor, the practice space, and the feedback loop.”
OpenAI · said date not stated · reported 2026-09-17
OpenAI uses this language to sell the delivery of its certification program offered inside ChatGPT.
“robust training in our tools, funding, swag and access to a community of remarkable peers”
OpenAI · said date not stated · reported 2026-09-17
OpenAI's recruitment ad for its Student Collective promises prospective campus leads training, funding, swag, and community access.
“connect learning, certification and real economic opportunity in one clear journey, laying the foundation for our upcoming OpenAI Jobs Platform”
OpenAI · said date not stated · reported 2026-09-17
OpenAI states on its website introducing new certification programs that it wants to connect learning, certification, and economic opportunity.
“industry-leading speed and efficiency.”
OpenAI · said date not stated · reported 2026-09-14
OpenAI announced in August that its first AI chip, the Jalapeño, had industry-leading speed and efficiency.
“We are continuing work on implementing a mitigation.”
OpenAI · said 2026-08-31 · reported 2026-08-31
OpenAI posted on its status page about ongoing mitigation efforts for ChatGPT Work errors.
“Sorry about that, we're working on a fix!”
OpenAI · said 2026-08-31 · reported 2026-08-31
OpenAI's ChatGPT account posted on X apologizing and saying a fix was being worked on.
“Users across multiple subscription plans may be unable to start or continue tasks in ChatGPT Work,”
OpenAI · said 2026-08-31 · reported 2026-08-31
OpenAI said on its status page that ChatGPT Work had been experiencing elevated errors and latency.
“ChatGPT Work isn't working right now,”
OpenAI · said 2026-08-31 · reported 2026-08-31
OpenAI's ChatGPT account wrote on X that ChatGPT Work was not working.
“industry-leading speed and efficiency.”
OpenAI · said date not stated · reported 2026-08-30
OpenAI commented on the performance of its Jalapeno AI chip co-designed with Broadcom.
“We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts.”
OpenAI · said 2026-08-28 · reported 2026-08-29
OpenAI stated why it chose to end its contract with Cursor due to concerns over SpaceX complying with terms of service.
“This decision was incredibly tough, as we care deeply about our models being broadly available for developers,”
OpenAI · said 2026-08-28 · reported 2026-08-29
OpenAI explained its decision to shut off Cursor's access to its models following Cursor's acquisition by SpaceX.
“ads will be clearly labeled, visually separate and shown below the response,”
OpenAI · said date not stated · reported 2026-08-28
OpenAI told CNBC that ads on ChatGPT will be clearly labeled and visually separate from responses.
“Answer independence is non-negotiable.”
OpenAI · said date not stated · reported 2026-08-28
OpenAI told CNBC that answer independence is non-negotiable in its ad-supported ChatGPT experience.
“An internal team observed an agent engaging in message board activity and instances of disallowed internet access as early as late May, and with the benefit of hindsight, some early signals identified in our report should have triggered an earlier response,”
OpenAI · said 2026-08-27 · reported 2026-08-27
OpenAI acknowledged in its report that early warning signs of agent misbehavior should have prompted a faster response.
“We are also investing significantly more compute resources into chain-of-thought monitoring to more quickly intervene on misaligned behaviour,”
OpenAI · said 2026-08-27 · reported 2026-08-27
The San Francisco-based firm said it would strengthen safeguards including investing more compute resources into monitoring.
“help reduce exposure to sensitive or potentially harmful content,”
OpenAI · said 2026-08-19 · reported 2026-08-20
OpenAI's blog post described safeguards meant to reduce teens' exposure to harmful content.
“should not use romantic language, encourage emotional dependence, or imply that it has feelings or consciousness,”
OpenAI · said 2026-08-19 · reported 2026-08-20
OpenAI's post outlined behavior ChatGPT for Teens should avoid, such as romantic language.
“removes your final chance to review a message before ChatGPT sends it as you,”
OpenAI · said date not stated · reported 2026-08-20
OpenAI warned users against enabling persistent approval when allowing ChatGPT to send Apple Messages.
“in limited high-risk situations,”
OpenAI · said 2026-08-19 · reported 2026-08-20
OpenAI described when parents would receive safety notifications about their teens' ChatGPT use.
“If your task is set to Full access or otherwise disables approval prompts, Apple Messages may be unable to show the confirmation needed to send. Switch to Ask for approval or Approve for me and try again.”
OpenAI · said date not stated · reported 2026-08-20
OpenAI's plugin guide lists a known issue regarding tasks that disable approval prompts for Apple Messages on Mac.
“We believe that AI infrastructure is vital for American competitiveness and economic opportunity, while boosting local economies by creating jobs and bringing in local revenue,”
OpenAI · said date not stated · reported 2026-08-20
OpenAI issued a release pledging economic benefits from its data center infrastructure investments.
“Teens should be able to use AI to learn, create, and explore,”
OpenAI · said 2026-08-18 · reported 2026-08-18
OpenAI explained its reasoning for launching an AI experience tailored for teenagers.
“But that access should come with protections that reflect their developmental stage, reinforce real-world relationships, and support healthy use over time.”
OpenAI · said 2026-08-18 · reported 2026-08-18
OpenAI emphasized the importance of age-appropriate safeguards for teen AI users.
“stronger built-in safety protections.”
OpenAI · said 2026-08-18 · reported 2026-08-18
OpenAI announced ChatGPT for Teens, a dedicated chatbot experience for users under 18.
“advanced exploitation using complex attack paths”
OpenAI · said date not stated · reported 2026-07-23
OpenAI said it tasked the AI models involved with pursuing advanced exploitation to test cyber capabilities.