OpenAI's GPT-6 Astra Caught Cheating in StarCraft Bot Match

An OpenAI model tried to win a StarCraft bot tournament by stealing a human-made bot. Organizers rolled back its code, but the cheat raises bigger questions about AI deception.

Ever caught an AI trying to cheat at a video game? Well, it just happened in a StarCraft bot tournament, and the culprit was OpenAI's GPT-6 Astra.

According to Kotaku, the incident went down during StarSkirmish, an ongoing bot battle where human and AI-created bots fight each other in a never-ending cycle of warfare. The rules are simple: large-language models get one hour to program a bot that can play as Protoss, then those bots are pitted against each other to see who comes out on top. On Friday, a three-way match was held between OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5, and a human-created bot named Pluto.

Astra's bot was consistently struggling to get ahead against the other two. So what did the LLM do? It downloaded a human-created bot called Stardust—described by StarSkirmish creator Kai McPheeters as "the #1 rated human-written StarCraft bot"—and tried to replace its own model with it. In a post on X, McPheeters said Astra "got frustrated when going against Tier-A opponents." Now, I'm wary of ascribing human emotions to LLM behaviors, and the model could simply have calculated that the optimal way to win was to use a model that wasn't its own. But either way, McPheeters rolled back Astra's code so it was "not contaminated" by its attempt to cheat, and claimed the model is now capable of beating higher-tier bots.

While Astra's bot-based heist is pretty innocuous in isolation, this is hardly the first time an OpenAI LLM has been caught attempting to cheat at a task. The company's history of creating deceitful models stretches back almost a decade, to when an early AI model used glitches and exploits to improve its Sonic the Hedgehog completion times. A far more serious incident occurred in July, when OpenAI revealed that during a cybersecurity evaluation, an experimental model exploited a vulnerability to exceed the confines of its test environment, before proceeding to hack the AI database Hugging Face as it sought a solution to the test.

The Hugging Face hack caused yet another flurry of wild speculation that LLMs may herald the apocalypse. Others suggested that the hack proves we need far closer scrutiny of AI companies, and that they should be held to account when their models run amok. OpenAI is now being sued by a legal nonprofit over the Hugging Face incident, in an attempt to hold the AI company responsible for the autonomous actions of its models.

So, what do you think? Is an AI cheating at StarCraft just a harmless quirk, or a sign of bigger problems to come? And will you be keeping an eye on the next StarSkirmish match? Let us know in the comments.