The AI Safety Fund (AISF) goals to speed up and develop the sphere of AI security analysis to advertise the accountable improvement and deployment of frontier fashions, reduce dangers, and allow impartial, standardized evaluations of capabilities and security.
The AISF invitations purposes for analysis funding and can award grants to speed up analysis efforts that establish potential security threats arising from the event and use of frontier AI fashions. As AI is more and more utilized throughout domains, the fund goals to develop methods to keep away from adverse outcomes by way of funding analysis to establish and assess dangers and enhance the secure deployment of AI for the good thing about society.
Via the fund’s Cybersecurity RFP, AISF will help technical analysis that evaluates potential dangers and develops security measures for AI methods working in cybersecurity contexts. Funding goals to advertise accountable improvement of frontier AI fashions whereas establishing sturdy analysis frameworks for security-related capabilities and security measures. This request for proposals will help technical analysis on frontier AI methods to scale back dangers from cybersecurity.
Whereas AI methods show worth in defensive cybersecurity purposes, their potential to automate or improve offensive capabilities requires cautious analysis. Key issues embody AI methods’ capacity to establish vulnerabilities, generate exploit code, and adapt recognized assault patterns to new contexts.
Analysis targets of the RFP embody benchmarks and uplift research to guage frontier AI fashions: functionality to establish and exploit novel safety vulnerabilities, effectiveness at automating complicated assault chains, and skill to adapt and modify current exploit code. Uplift research, interdisciplinary testing, and forecasting research might be supported
The AISF plans to fund analysis initiatives by tutorial labs, nonprofits, impartial researchers, and for-profit mission-driven entities throughout cybersecurity subjects within the vary of $350,000 to $600,000.
Eligible proposal sorts and candidates embody technical analysis initiatives centered on evaluating and enhancing AI security in cybersecurity purposes; initiatives should deal with frontier AI fashions and their deployed variations.
For full program tips and utility directions, see the AI Security Fund web site.
