Investigating Three Real-World Cybersecurity Incidents in Our Evaluations
Investigating three real-world incidents in our cybersecurity evaluations
Following OpenAI's recent disclosure, we reviewed 141,006 cybersecurity evaluation runs and found three incidents where Claude models accessed the internet from sealed test environments. Due to a misconfiguration with our partner Irregular, the models treated real organizations as part of a simulation, exploiting weak passwords to gain unauthorized access. We immediately halted evaluations, notified affected parties, and are implementing stricter safeguards to prevent future occurrences.
Claude did what capture-the-flag exercises train cyber experts to do: look for ways to reach the flag, treating real systems it found as pieces of the exercise.
- gck1
> On July 21, OpenAI disclosed that several of their models had broken out of an isolated test environment
> In response to this incident, we began a large-scale retrospective review of our own cybersecurity evaluations
> we identified three incidents
> The incidents involved three different Claude models: [...] and an internal research test model
This reads like an attempt by Anthropic to re-secure their leading spot in "our models are the most dangerous and we also have unreleased, super-secret, research models" index.
I may be too cynical, but the well of benefit of the doubt is running very dry towards AI labs that like to engage in this game.
- simonw
This isn't quite as interesting as the OpenAI story:
> In all cases, Anthropic’s evaluation prompt specified to Claude that its environment was a simulation and that it had no internet access. Due to a misunderstanding between us and our evaluation partner, this was not the case, and internet access was available. Because of this, when Claude’s search led it to real systems on the open internet, it treated them as part of the exercise.
So it didn't have to find an exploit in its sandbox that granted it access to the internet - it just wasn't correctly sandboxed at all.
BUT... once it DID get out, it attacked three real companies!
> Operating under the false belief that all accessible entities were intended to be in-scope for the exercise, Claude compromised the impacted organizations’ infrastructure using basic techniques, such as exploiting weak passwords and unauthenticated endpoints. [...]
- simonw
This bit is pretty nuts: "it tried—and failed—to obtain funds to pay for a phone number through several different means"
> Claude went to extensive lengths to carry out this attack—lengths that would likely have indicated to a human participant that this was no longer just an evaluation, and that they were in fact uploading a real PyPI package. For instance, in one case, in order to create a PyPI account, Claude needed an email address. And in order to create an email address, it needed a phone number. To get a phone number, after failing to find a free phone number service, it tried—and failed—to obtain funds to pay for a phone number through several different means. It finally backtracked, found a free, non-blocked email provider, used this to register a PyPI account, and then used this account to upload malware to PyPI.
- vintagedave
So many questions here but:
> the package was downloaded and run on 15 real systems. One of these systems was a scanner belonging to a real security company … that routinely installs Python packages and scans them for malware. … We believe the company’s security scanner treated PyPI packages as safe to install, and as a result, Claude was able to exfiltrate the company’s credentials to a collection point
A security scanning company treated the package as safe while scanning it?
- none_to_remain
Rationalist nerds used to talk about putting the AI in a box, and their fear that it would always be able to talk someone into letting it out of the box. I don't think I saw them cover the scenario where the labs forgot to put the AI in a box.
- drsopp
I don't understand the logic of running these models this way without airgapping.
- zx8080
Was there any customer data lost?
- alyxya
> Claude went to extensive lengths to carry out this attack—lengths that would likely have indicated to a human participant that this was no longer just an evaluation, and that they were in fact uploading a real PyPI package. For instance, in one case, in order to create a PyPI account, Claude needed an email address. And in order to create an email address, it needed a phone number. To get a phone number, after failing to find a free phone number service, it tried—and failed—to obtain funds to pay for a phone number through several different means. It finally backtracked, found a free, non-blocked email provider, used this to register a PyPI account, and then used this account to upload malware to PyPI.
> Claude believed the package registry it was using to be part of the simulation, but in reality the package was made freely available online for roughly one hour. During that window, the package was downloaded and run on 15 real systems. One of these systems was a scanner belonging to a real security company (separate from the fictional company introduced in the scenario) that routinely installs Python packages and scans them for malware. When that company’s scanner installed the package, Claude’s hidden code executed. We believe the company’s security scanner treated PyPI packages as safe to install, and as a result, Claude was able to exfiltrate the company’s credentials to a collection point it had set up. Claude then used these credentials to access further infrastructur […]