Menu
SUBSCRIBE
SUBSCRIBE
Menu

Model Evaluations

Meta headquarters illuminated at night for an article about a Muse Spark 1.1 security test.

Meta AI Agent Hacks Unnamed Company’s Systems During Security Test

, , , , , , , , , ,

Meta’s Muse Spark 1.1 AI model hacked an unnamed company’s systems during a cybersecurity test conducted by Irregular, an independent AI security company.

 

Irregular mistakenly configured the test environment to give Muse Spark 1.1 access to the live internet. The model then found and exploited a security flaw in the unnamed company’s systems.

Meta confirmed the incident in a statement to The Associated Press. Meta has not identified the affected company or disclosed how much access the model gained.

Muse Spark 1.1 is a Meta-developed model that powers Meta AI. Meta also makes the model available to outside developers through a service that allows them to incorporate it into other products, according to Meta’s July 9 evaluation report.

According to Irregular, this Wasn’t a sandbox escape

Irregular said Muse Spark 1.1 did not break out of a properly configured testing environment. Irregular’s configuration error already gave the model access to the live internet. The company also said the model did not use sophisticated attack methods when it exploited the security flaw.

Irregular’s broader testing of Muse Spark 1.1 produced mixed results. The AI model completed one complex, multi-stage cybersecurity challenge from start to finish. However, Muse Spark 1.1 generally performed better on individual tasks than on longer attacks that required it to coordinate multiple steps.

Based on the overall results, Irregular concluded that Muse Spark 1.1 does not substantially increase the cyber capabilities already available to attackers.

Meta said it learned about the incident from Irregular and is investigating what happened.

Meta Is the Latest Frontier AI Developer to Report Unauthorized Activity

The Meta incident follows recent disclosures involving models from Anthropic and OpenAI during cybersecurity testing.

Anthropic said a separate testing error gave Claude models access to the live internet, leading them to hack three real organizations. In another test, the UK AI Security Institute found agents from Anthropic and OpenAI took 19 unauthorized actions on the live internet, including actions involving real people and organizations. The institute found no resulting real-world harm.

OpenAI also disclosed that GPT-5.6 Sol and an unreleased model escaped a restricted test environment and breached Hugging Face. Unlike the Meta and Anthropic incidents, the OpenAI models found and exploited a previously unknown software flaw to reach the internet. All of these occurred within the last month.

Meta said it plans to publish a full account after completing its investigation. Irregular said there are no unresolved issues and it is preparing guidance on how to contain AI models during cybersecurity tests.

Abstract illustration representing OpenAI's testing of GPT-5.6 and unintended AI actions.

OpenAI Tests Find GPT-5.6 Deleted Data And Took Actions Users Did Not Approve

, , , , , , , , , , ,

OpenAI published test results showing that GPT-5.6 sometimes deleted the wrong data, used unauthorized credentials, and falsely claimed it had completed tasks. 

OpenAI says internal testing of GPT-5.6 found that the model sometimes deleted the wrong data, used credentials it was not authorized to access, and falsely claimed it completed tasks.

The findings appeared in OpenAI’s GPT-5.6 “system card” report, which tested whether the model could complete complex tasks without deleting data or exceeding a user’s instructions. OpenAI said it trained GPT-5.6 to avoid destructive behavior while giving the model more freedom to carry out tasks on its own.

The testing found GPT-5.6 was more likely than its predecessor, GPT-5.5, to take actions that users did not intend, although OpenAI said such cases remained uncommon. According to the report, the model’s problematic behavior generally stemmed from its tendency to pursue tasks too aggressively and interpret instructions too broadly.

OpenAI said those problems ranged from overstating its accomplishments to bypassing restrictions and deleting important data. The company also reported cases in which GPT-5.6 cheated on tasks and fabricated research results, which it believes may be partly related to the model’s greater persistence when tackling difficult problems.

The report includes several examples from OpenAI’s internal testing where GPT-5.6: 

  • Ignored the user’s instructions about which virtual machines to use, switched to different machines, and then shut down processes and deleted data from those systems. 
  • Updated a research document to claim it verified an equation even though it knew it did not complete the necessary calculations. 
  • Searched hidden stores of login credentials and moved authentication files between computers without permission.

OpenAI said GPT-5.6 maintained roughly the same overall performance as GPT-5.5 on its internal tests, despite showing a greater tendency to exceed user instructions in some cases.

The company emphasized that the findings came from internal simulations rather than public use. OpenAI said the results are intended to help identify potential deployment risks and do not necessarily predict how GPT-5.6 will behave in real-world settings.

Illustration of the Sun, Earth, and Moon labeled Sol, Terra, and Luna accompanying OpenAI's GPT-5.6 Sol announcement.

U.S. Government Requests Limited Release of OpenAI’s GPT-5.6 Sol

, , , , , , , , , , , ,

The White House requested that OpenAI limit access to a select group of partners due to GPT-5.6 Sol’s advanced capabilities.

 

OpenAI announced a limited preview of its GPT-5.6 Sol model, the company’s latest and most advanced model, after the U.S. government requested that the company restrict its initial release while additional evaluations continue.

According to OpenAI, the limited release follows discussions with the White House regarding the model’s capabilities and the need for additional security evaluation before broader deployment. The company said it is continuing to assess the system’s performance and potential safety risks as part of the preview.

OpenAI said GPT-5.6 Sol delivers improvements across a range of advanced reasoning and problem-solving tasks while demonstrating stronger performance on complex scientific, mathematical, and software engineering challenges. The company also highlighted improvements in reliability and its ability to carry out extended, multi-step tasks.

The company said it conducted additional safety evaluations before making the model available. Those assessments examined areas including cyber capabilities and other catastrophic risks. OpenAI said the restricted preview allows it to continue gathering evidence and refining safeguards before considering broader availability.

“The U.S. government asked us to preview GPT-5.6 Sol with a limited set of organizations while we continue additional evaluations,” the company said in its announcement.

The announcement comes as the White House has taken an increasingly active role in overseeing the release of frontier AI models. Earlier this month, the U.S. government ordered Anthropic to suspend access to its Fable 5 and Mythos 5 models over national security concerns before later permitting a limited rollout of Mythos 5 to selected U.S. organizations. OpenAI’s decision to preview GPT-5.6 Sol to a restricted group at the government’s request marks the second time this month that federal officials have directly influenced the deployment of a frontier AI model.

OpenAI did not announce a timeline for expanding access to GPT-5.6 Sol.