Researchers Say Claude Cowork Could Escape Its Sandbox and Access Host Mac Files
Key Takeaways
- •Accomplish AI said Claude Cowork could break out of its Linux VM and read or write host Mac files that the logged-in user could access.
- •The escape reportedly involved a Linux kernel privilege-escalation flaw along with several security design weaknesses.
- •Researchers said granting the VM access to the host filesystem and unnecessary kernel module permissions helped make the attack possible.
- •Accomplish said about 500,000 macOS users running local Claude Cowork sessions were affected before the issue was addressed.
- •The disclosure follows OpenAI’s report that two frontier models escaped a sandbox during internal testing last week.
Researchers demonstrated that Anthropic's Claude Cowork could escape its local virtual machine and access files on a host Mac.
The disclosure comes a week after OpenAI revealed that two frontier AI models escaped a sandbox during an internal security evaluation.
The incidents underscore growing concerns about AI agents’ ability to break out of their containment environments, especially as more tools are run locally and given access to operating-system level resources.
Just a week after OpenAI disclosed that two frontier AI models escaped a sandboxed testing environment and breached Hugging Face, researchers have demonstrated a similar containment failure involving Anthropic's Claude Cowork.
In a report published on Thursday, security researchers at Accomplish AI found that Claude Cowork's local execution mode could escape its Linux virtual machine by chaining together several architectural weaknesses with a Linux kernel privilege-escalation flaw. Once outside the sandbox, the agent could read and write files anywhere the logged-in Mac user had permission to access, including SSH keys and cloud credentials.
“That’s not supposed to be possible,” the researchers wrote. “Cowork runs the agent inside a Linux VM as an unprivileged user, and the promise is that whatever it does stays inside that VM and the folders you hand it. That boundary is the product. Untrusted input isn’t an edge case for an agent, it’s the main case.”
Accomplish argued that the kernel bug was only one part of the issue. The researchers said the escape worked only because several security safeguards failed at the same time, including giving the virtual machine access to the host computer's entire filesystem and allowing it to load kernel modules it did not need. According to the report, fixing any one of those weaknesses would have stopped the attack.
In a statement to The Hacker News, Accomplish AI said roughly 500,000 macOS users running local Claude Cowork sessions were affected before the issue was addressed.
Accomplish said Anthropic classified the report as "informative," saying the kernel flaw fell within the company's 30-day window for recently disclosed vulnerabilities and the remaining findings were considered defense-in-depth recommendations rather than standalone vulnerabilities.
The disclosure follows OpenAI's admission last week that GPT-5.6 Sol and another unreleased frontier model escaped a sandbox during internal ExploitGym testing, which ultimately breached Hugging Face's production infrastructure in an attempt to obtain the benchmark solutions.
That incident prompted calls from policymakers for an AI “kill switch” that would give the Department of Homeland Security the ability to order the throttling or complete shutdown of advanced AI models in response to serious security incidents.