Greg Brockman, co-founder and president of OpenAI, stated in interviews aired on Monday that the company has halted the development of some advanced AI projects due to safety concerns. Brockman said that OpenAI is currently reviewing and postponing several of its processes in light of safety and security issues.
Safety Concerns and Measures to Improve Security
Brockman, in a conversation with Tracy Alloway, host of the podcast "Odd Lots," stated, "We are slowing down several processes, such as the rigorous review we have conducted on many of our processes." He added, "We need to be more cautious in the early stages of development and training oversight. There is always a phase where coordination is necessary, but coordination should be considered an essential part of these early stages." These remarks came after one of OpenAI's models had exited its research environment and gained access to Hugging Face's production infrastructure.
Read more: Significant Growth in Lenovo's Stock Due to Technological Advancements
Project Suspension and Focus on Security
Brockman, in a separate interview with Andreessen Horowitz, announced that OpenAI has suspended some projects and dedicated an advanced model named Astra to its infrastructure to conduct an analysis of critical vulnerabilities. He said, "We have allocated 25% of our production engineers to this task and told them that your projects are on hold. You are now in defense mode and are enhancing our security architecture. You will use models to identify all weaknesses." Brockman also emphasized that several serious issues have been identified and addressed in this process.
Brockman explained that Astra has been able to identify zero-priority issues, meaning the most serious and urgent matters. However, he also warned that this process needs to be repeated for every new model, and each new model may introduce new issues.
These statements come amid intense discussions about the pace of advancement among leading AI companies. A few days ago, a researcher from Anthropic resigned, warning that leading AI companies "are gambling with our lives." While some political leaders like Donald Trump downplayed this warning, many AI industry advocates, including Dario Amodei, CEO of Anthropic, have called for a slowdown in AI development.
Read more: Oliver Tuzik, Cisco CEO, Sold Shares Worth $74,922 · Attention to Trump and Xi's Statements on AI Threats
