Google announces new AI Cyber Defense Initiative
Google has confirmed a new security scheme which, it says, will help “secure, empower and advance our collective digital future” using AI. Part of this AI Cyber Defence Initiative includes open-sourcing the new, AI-powered, Magika tool that is already being used to help protect Gmail users from potentially problematic content.
The AI Cyber Defense Initiative, to be launched at the Munich Security Conference on February 16, signals a belief that Google’s experience in deploying AI can help to reverse what it calls the defender’s dilemma. This can be defined as the process by which attackers only need to find one hole, one novel threat, in order to break down and pierce even the best-defended of networks, whereas defenders have no margin for error and must deploy the best defenses at all times. This is perhaps best summed up by realizing that preventative measures, patches, and even public awareness campaigns are all geared to mitigate yesterday’s threats rather than those, as Google puts it, that are waiting in the wings.
Google will publish a new report during the Munich conference, Secure, Empower, Advance: How AI Can Reverse the Defender’s Dilemma, which will highlight the newly proposed policy and technology agenda. Against a backdrop of extensive experience in deploying AI at scale, Google will argue that the always-in-catch-up mode defender’s dynamic can be reversed. This is because, Phil Venables, chief information security officer for Google Cloud, and Royal Hansen, vice-president of engineering for privacy, safety and security at Google, say, AI enables security professionals to “scale their work in threat detection, malware analysis, vulnerability detection, vulnerability fixing and incident response.”
Gmail is already using a technology by the name of RETVec, a multilingual neuro-based text processing model that has led to spam-detection rates improving by 40% while, at the same time, reducing the false positive rate by 19%. Given that spam is a primary carrier of malicious payloads this should not be dismissed lightly.
Gmail also uses a new powered by AI tool called Magika to help with file type identification in order to help defenders detect malware. Google is to open-source Magika, which is also used to help protect Google Drive and the Safe Browsing functionality, thus increasing its footprint and success in thwarting attackers.
According to Google, Magika is already outperforming standard, conventional, file identification methodology by as much as 30% when it comes to accuracy. When it comes to the hardest to identify of potentially problematic content, that which uses JavaScript, Powershell or VBA, for example, the precision rate climbs to a staggering 95%.
The AI Cyber Defense Initiative will see Google making new commitments to invest in AI-ready infrastructure, as well as releasing release new tools for defenders, alongside new research and security training. A ‘Secure AI Framework’ is being launched to aid best practice collaboration when it comes to AI system security. This is in addition to some 17 startups spread across the U.S., U.K., and Europe as part of the Google for Startups Growth Academy’s AI for Cybersecurity Program. Once again, collaboration is key and the aim is for the AI for Cybersecurity program to help build and strengthen a transatlantic ecosystem for cybersecurity that includes AI tools and the skills needed to effectively use them.
Google is also announcing $2 million in research grants to help advance breakthroughs in AI-powered security including code verification enhancement and the development of large language models that can be more resilient to threats. The funding will be supporting researchers at the University of Chicago, Carnegie Mellon, and Stanford to begin with.

source