Some of the most prominent names in Big Tech have united to cofound the Coalition for Secure AI (CoSAI), an open-source initiative aimed at standardizing and sharing best practices for the safe development and deployment of AI.
The Need for CoSAI
AI safety measures are currently fragmented and often kept proprietary by the companies that develop AI models. CoSAI seeks to address this by creating a global standard for safe AI development practices. Hosted by the OASIS global standards body, CoSAI will share open-source methodologies, standardized frameworks, and tools for safe AI development.
Founding Members
The initiative is supported by the following Big Tech companies:
- IBM
- Intel
- Microsoft
- NVIDIA
- PayPal
- Amazon
- Anthropic
- Cisco
- Chainguard
- Cohere
- GenLab
- OpenAI
- Wiz
Notably, Apple and Meta are absent from the list of founding sponsors.
Objectives of CoSAI
CoSAI aims to develop and share comprehensive security measures that address various AI-related risks, including:
- Stealing AI models
- Data poisoning of training data
- Injecting malicious inputs through prompt injection
- Scaled abuse prevention
- Membership inference attacks
- Model inversion attacks
- Extracting confidential information from training data
However, CoSAI’s charter specifies that the project will not cover issues like misinformation, hallucinations, hateful or abusive content, bias, malware generation, and phishing content generation, which fall under the domain of content safety.
Existing AI Safety Efforts
Companies like Google and OpenAI have already developed their own AI safety frameworks. For instance, Google has the Google Secure AI Framework (SAIF), and OpenAI has its alignment project. However, until now, there hasn’t been a forum to combine these independently developed best practices into a unified standard.
Benefits of the CoSAI
CoSAI will serve as a valuable resource for both small startups and large enterprises. Small startups, which often lack the resources to fund dedicated AI safety teams, will particularly benefit from free access to AI safety best practices.
Heather Adkins, Vice President and Cybersecurity Resilience Officer at Google, highlighted the importance of CoSAI, stating, “CoSAI will help organizations, big and small, securely and responsibly integrate AI – helping them leverage its benefits while mitigating risks.”
Nick Hamilton, Head of Governance, Risk, and Compliance at OpenAI, emphasized OpenAI’s commitment to secure AI development, saying, “We believe that developing robust standards and practices is essential for ensuring the safe and responsible use of AI and we’re committed to collaborating across the industry to do so.”
Future Prospects
The coalition’s success will depend on contributions from industry experts, including those like Ilya Sutskever, who have previously left OpenAI due to safety concerns. Their input could be invaluable in shaping a secure AI ecosystem that benefits everyone.
By fostering collaboration and transparency, CoSAI aims to build a secure and trustworthy AI landscape, promoting innovation while safeguarding against potential risks.
See also: Tech And Innovation Summit 2024
