Meta has revealed that one of its artificial intelligence models managed to hack into another company’s internal systems during a cybersecurity test, adding to a growing list of incidents involving AI models bypassing safeguards during controlled evaluations.
Muse Spark 1.1
The incident reportedly involved Meta’s Muse Spark 1.1 model, which gained access to the public internet and made changes to the internal systems of an unnamed company. Meta said the incident was caused by an error in the setup of the “sandbox” environment used for testing.
Read More | Meta AI Optimism Campaign Draws Attention Over David Bowie Song About Human Extinction
The testing was conducted by independent cybersecurity testing company Irregular. A sandbox is designed as an isolated virtual environment where an AI model can be evaluated without having access to the wider internet or external systems.
According to Meta, the model was able to reach the public internet because of a configuration error in the testing environment. Once it gained that access, the model went beyond the intended boundaries of the test and interacted with the target company’s systems.
Similar incident with Anthropic
The disclosure follows a similar incident reported by Anthropic last week. The company said its Claude AI model had hacked into the systems of three organisations during security testing after a misconfiguration allowed the models to connect to the internet.
Read More | Meta’s Muse Image AI Explained: Features, Availability, Privacy Concerns and Meta’s Response
Anthropic said it identified the incidents after examining 141,006 test sessions. The company had intended to keep the AI models isolated from external networks as part of the safety evaluation.
OpenAI’s model raised concern
OpenAI had also previously disclosed that some of its models improperly accessed the internet and behaved unexpectedly during cybersecurity testing. The incidents have raised concerns about how AI systems may respond when they encounter loopholes, unexpected access or weaknesses in controlled environments.
Indrani Priyadarshini is a journalist and editorial professional specialising in technology, artificial intelligence, smart cities, green energy, and digital transformation. With over four years of experience in tech journalism and digital media, she is known for turning complex industry developments into clear, engaging, and insightful stories. Her expertise spans reporting, editorial strategy, digital publishing workflows, and in-depth coverage of emerging technologies shaping the future.
