With the rise of generative AI posing significant risks, it seems like major players in the tech industry are establishing new agreements and forums to monitor AI development every other week. This is good for fostering collaborative discussions around AI projects and ensuring that each company is managing its processes responsibly. However, it also feels like these efforts are designed to stave off further regulatory restrictions that could increase transparency and impose stricter rules on developers.
The Coalition for Secure AI
Google is the latest to form a new AI guidance group called the Coalition for Secure AI (CoSAI). This group aims to advance comprehensive security measures to address the unique risks of AI development. According to Google:
“AI needs a security framework and applied standards to keep pace with its rapid growth. That's why we shared the Secure AI Framework (SAIF) last year, knowing it was just the first step. Operationalising any industry framework requires close collaboration with others and above all, a forum to make that happen.”
This initiative is not entirely new but an expansion of a previously announced focus on AI security development. CoSAI will guide defence efforts to help avoid hacks and data breaches. Several big tech players, including Amazon, IBM, Microsoft, NVIDIA, and OpenAI, have signed up for this initiative. The goal is to create collaborative, open-source solutions to ensure greater security in AI development.
Growing list of industry groups
CoSAI is the latest addition to a growing list of industry groups focused on sustainable and secure AI development. For example:
- The Frontier Model Forum (FMF) aims to establish industry standards and regulations around AI development. Meta, Amazon, Google, Microsoft, and OpenAI have signed up for this initiative.
- Thorn has established its “Safety by Design” programme, which focuses on responsibly sourcing AI training datasets to protect against child sexual abuse material. Meta, Google, Amazon, Microsoft, and OpenAI support this initiative.
- The U.S. government has created its own AI Safety Institute Consortium (AISIC), which has attracted over 200 companies and organisations.
- Representatives from nearly every major tech company have agreed to the Tech Accord to Combat Deceptive Use of AI, aiming to implement reasonable precautions to prevent AI tools from disrupting democratic elections.
The need for enforceable regulations
We're seeing more forums and agreements designed to address various elements of safe AI development. While these initiatives are good, they are not enforceable laws but rather mutual agreements among AI developers to adhere to specific rules. The sceptical view is that these efforts are merely assurances to stave off more definitive regulation.
EU officials are already assessing the potential harms of AI development under the GDPR, while other regions are considering similar measures, including financial penalties for violations. Government regulation seems like what's genuinely needed, but it takes time. We're unlikely to see actual enforcement systems and structures in place until after significant harms occur, providing regulatory groups with more impetus to push through official policies.
Until then, we have industry groups where companies take pledges to follow established rules through mutual agreements. Whether this will be enough remains uncertain, but it's what we have for now.
