Several influential artificial intelligence (AI) companies have teamed up to prioritize security in the development and use of generative AI. The Coalition for Secure AI, known as CoSAI, is working towards establishing a security-first approach to AI to address the risks involved in its deployment. The main objective of CoSAI is to create standardized guardrails, security technologies, and tools for the secure development of AI models.

According to a statement from CoSAI, the initial focus areas include AI and software supply chain security, as well as preparing defenders for the evolving cyber landscape. The coalition’s founding members, which include Google, OpenAI, and Anthropic, are leading efforts to create a secure environment around the access and use of AI models. The goal is to shield AI models from potential cyberattacks and ensure their integrity.

Google, a key member of CoSAI, highlighted the importance of establishing a framework for AI security that is both timely and responsible. Heather Adkins, Google’s vice president of security engineering, and Phil Venables, Google Cloud’s chief information security officer, emphasized the need for continuous updates and enhancements to meet the evolving cybersecurity challenges.

AI safety has become a top priority in the tech industry, especially following the launch of ChatGPT in 2022. Concerns over cybersecurity threats related to AI misuse for social engineering and the creation of deepfake content have been on the rise. Security firms like Trend Micro and CrowdStrike are now leveraging AI to help companies detect and mitigate potential threats.

Gartner analyst Avivah Litan emphasized the importance of ensuring AI safety, trust, and transparency to avoid detrimental outcomes for organizations. Litan pointed out that AI must operate within defined parameters and guidelines to prevent errors and exceptions that could lead to negative consequences. The emergence of AI agents, which enhance data processing, has added complexity to AI security challenges, requiring robust tools for effective remediation.

US President Joe Biden has underscored the significance of AI safety and ethics in safeguarding national security and promoting equity. In July 2023, President Biden issued an executive order calling on major companies associated with CoSAI to adhere to safety standards, share test results, and prevent AI misuse for nefarious purposes such as fraud.

CoSAI aims to collaborate with various organizations, including the Frontier Model Forum, Partnership on AI, OpenSSF, and MLCommons, to develop common standards and best practices for AI security. MLCommons is set to launch an AI safety benchmarking suite that will assess the performance of large language models in handling sensitive topics like hate speech, exploitation, and abuse.

The management of CoSAI will fall under OASIS Open, an organization known for its work on open-source projects. OASIS, akin to the Linux Foundation, has a track record of overseeing standardization efforts, including the XML standard and the ODF file format. Through these collective efforts, CoSAI and its partners are committed to promoting a security-first approach to AI development and usage.

Source link