Amid rising concerns around the security of artificial intelligence models, Meta has moved to improve the integrity of its systems by hiring more talent with specialized skills in this area.
As reported by Axios, Meta hired three founders from AI security startup Virtue AI in order to improve internal testing and ensure safety within its AI models.
Virtue AI has established a reputation for stress-testing AI systems to detect potential vulnerabilities before bad actors uncover them first. The addition of these new hires to Meta’s team should help to make Meta’s systems more resilient to outside threats, while also offering assurances to vendors looking to implement these systems.
The move comes after the U.S. government recently issued an export control directive that restricts AI provider Anthropic from shipping its latest Fable 5 and Mythos 5 models to any foreign national group.
According to Anthropic, the government raised concerns about the potential misuse of Anthropic’s advanced models for cyber attacks. These latest models are able to bypass security protocols and exploit digital defenses in external systems.
Anthropic has implemented safeguards to reduce the usage of these tools for tasks related to cybersecurity. But the concern is that as AI models get stronger and more capable, they will be vulnerable to hackers and external groups who want to break into sensitive systems.
Which is why Meta’s looking to get ahead of this and develop a better understanding of the potential threats posed by its own AI models.
The focus on more advanced security measures will also help Meta ensure that its coming AI products meet the requirements of regulators in various regions, which are now looking for ways to ensure that AI is used safely, and doesn’t pose a public risk.
It’s hard to definitively say how dangerous AI models could be, because the complexity of these systems is still in flux. However, the addition of the Virtue AI team’s expertise could end up being a key step toward securing Meta’s future AI prospects.
And with Meta investing hundreds of billions into AI development, the company needs to also establish pathways to monetization, on many fronts, to offset that spending. Enhanced assurance will also be an important driver for this element.
