Introducing Google’s Safe AI Framework
The potential of AI, particularly generative AI, is immense. Nevertheless, within the pursuit of progress inside these new frontiers of innovation, there must be clear trade safety requirements for constructing and deploying this know-how in a accountable method. That’s why at this time we’re excited to introduce the Safe AI Framework (SAIF), a conceptual framework for safe AI programs.
- For a abstract of SAIF, click on this PDF.
- For examples of how practitioners can implement SAIF, click on this PDF.
Why we’re introducing SAIF now
SAIF is impressed by the safety greatest practices — like reviewing, testing and controlling the provision chain — that we’ve utilized to software program growth, whereas incorporating our understanding of security mega-trends and dangers particular to AI programs.
A framework throughout the private and non-private sectors is important for ensuring that accountable actors safeguard the know-how that helps AI developments, in order that when AI fashions are applied, they’re secure-by-default. As we speak marks an vital first step.
Over time at Google, we’ve embraced an open and collaborative approach to cybersecurity. This consists of combining frontline intelligence, experience, and innovation with a dedication to share risk info with others to assist reply to — and forestall — cyber assaults. Constructing on that strategy, SAIF is designed to assist mitigate dangers particular to AI programs like stealing the model, data poisoning of the training data, injecting malicious inputs by prompt injection, and extracting confidential information within the coaching information. As AI capabilities develop into more and more built-in into merchandise internationally, adhering to a bold and responsible framework shall be much more essential.
And with that, let’s check out SAIF and its six core components:
