OpenAI

"In June 2025, we proactively identified an account associated with this individual [Jesse Van Rootselaar] via our abuse detection and enforcement efforts, which include automated tools and human investigations to identify misuses of our models in furtherance of violent activities."

"The shooter evaded detection systems, but the second account was eventually discovered and shared with authorities."

"there is no evidence to suggest that user data was accessed, nor that its systems, intellectual property, or software were compromised or altered."

"when the right conditions, such as regulation and the cost of energy, enable long-term infrastructure investment."

"flexibility, certainty, and a focus on delivering the benefits of AI broadly."

"a single 'little goblin' in an answer could be harmless, even charming"

"it really isn't a marketing gimmick"

"Starting with GPT‑5.1, our models began developing a strange habit: they increasingly mentioned goblins, gremlins, and other creatures in their metaphors."

"The goblins were funny at first, but the increasing number of employee reports became concerning."

"We had a suspicion that something in our personality instruction-following training was amplifying this."

"Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query."