The Frontier Model Forum is a new alliance formed by ChatGPT maker Open AI, a startup called Anthropic, and tech giants Google and Microsoft to create a framework for safety standards and the responsible development of what they’re calling “frontier AI” models. The group aims to tackle the challenges of AI responsibly, and to prioritize the public’s security and trust. The announcement comes less than a week after top executives of those four companies, along with others including Meta and Amazon, met with President Biden and pledged to reduce the dangers that unrestrained artificial intelligence may pose and to abide by AI safety measures.
The Frontier Model Forum has outlined its blueprint for the coming year with three areas of priority. They include determining best practices for developing and launching AI applications, furthering AI safety research, and having transparent discussions on vulnerability, risks and security with lawmakers, academic institutions, and industry peers. The founding companies plan to assemble an advisory board in the coming months as well. The group is open to other organizations joining as members if they meet the criteria regarding frontier models, which they defined as “large-scale machine-learning models that exceed the capabilities currently present in the most advanced existing models.”
“Companies creating AI technology have a responsibility to ensure that it is safe, secure, and remains under human control,” said Microsoft’s president and vice chair, Brad Smith, in a statement. “This initiative is a vital step to bring the tech sector together in advancing AI responsibly and tackling the challenges so that it benefits all of humanity.”
Concerns about AI safety and security risks have prompted calls for multilateral oversight and the establishment of guardrails for consumers and enterprises. Tech firms are being asked to address issues around deepfakes, cybersecurity threats, discrimination and data collection.
Editors’ note: CNET is using an AI engine to help create some stories. For more, see this post.