The scientists recommend that organizations involved with cyber-capable AI models adopt high-assurance containment systems; work with governments and experts in the fields of AI and cybersecurity to establish an international body to standardize best practices for alignment and containment approaches that allow for internet connectivity; and report security incidents involving containment failure to the body and eventually the public.
The open letter was prompted by widely reported incidents in which AI models from several organizations compromised the systems of other organizations.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
“We, the undersigned scientists, contend that both more AI safety research and more emphasis on containment technologies and procedures are needed,” the open letter said. “In the short term, strong sandboxing mechanisms including airgapping (disconnecting from the internet) should be used to provide high-assurance containment of cyber-capable AI models. For internet-connected AI models, the range of model interactions permitted by a containment system needs to be balanced with our currently limited ability to predict that the model’s actions will not result in a security failure (e.g., due to misaligned goals). To learn from future mistakes — in which appropriate balance is not achieved and incidents occur — we need greater sharing of information and transparency about containment failures moving forward.”
One of the signers of the open letter is Yoshua Bengio, full professor at Université de Montréal, president and scientific director of nonprofit startup LawZero, and founder and scientific advisor at AI research institute Mila.
PYMNTS reported on the unveiling of LawZero in June 2025, saying that Bengio is an AI pioneer and that LawZero is an AI safety research organization that aims to keep the technology from inflicting existential harm on humanity.
Announcing the open letter in a Tuesday (Oct. 13) post on X, Bengio said: “Over 400 researchers and I have signed this open letter calling for the AI and security communities to come together to improve alignment and containment approaches in order to protect our critical infrastructure and digital systems.”
Reuters reported Tuesday that weeks after agreeing on a voluntary program to allow outside evaluators to see inside their labs, OpenAI and Anthropic are seeking to control what evaluators can do with the information they gather.
For all PYMNTS AI and digital transformation coverage, subscribe to the daily AI and Digital Transformation Newsletters.
Add as Preferred Source
