OpenAI stated that it will comply with the executive order signed by U.S. President Donald Trump a few days ago regarding pre-release review of artificial intelligence models, and agrees to accept a federal government capability assessment before the model is officially launched to the outside world.
George Osborne, OpenAI’s head of national affairs, confirmed in an interview with US financial media that the company will sign this voluntary administrative order and participate in the relevant evaluation process. "Democratic governments should play an important role in how this technology is used and deployed," he said.
Osborne made the remarks during an interview during the South by Southwest (SXSW) event in London. He emphasized that OpenAI takes its responsibilities "very seriously" and said: "As a leading laboratory with very powerful and cutting-edge artificial intelligence models, we will not wait passively for requests." He added that the company is "proactively making recommendations to governments to help them continue to track safety and security-related issues, not just in the United States, but also more broadly internationally."
Under the executive order signed by Trump on Tuesday, the federal government can gain access to artificial intelligence models 30 days before they are officially released. The order requires companies to participate in a benchmarking process to evaluate the "advancement" of models in terms of network capabilities and other aspects, and use this to determine when a model should be designated as a "regulated leading edge model."
The order is seen as a move by the Trump administration to balance cutting-edge artificial intelligence development with national security, and is intended to set a minimum regulatory threshold for the capabilities of large models without completely stifling innovation. In view of the potential military and security uses of large models in network attack and defense, intelligence analysis, automated attacks and other scenarios, how to define the regulatory boundaries of "cutting-edge models" is becoming a common issue for policymakers in various countries.
Osborne, who served as UK Chancellor of the Exchequer from 2010 to 2016, pointed out that governments “have to be smart enough” when regulating artificial intelligence. He said that one of OpenAI's recommendations to governments is to establish "strong but highly flexible" regulatory agencies so that these agencies can adjust regulatory methods and standards as technology develops.
Osborne's statement shows that in the context of intensified review pressure at the federal level in the United States, OpenAI has chosen to respond to supervision with a cooperative attitude rather than confrontation with the government. At a stage when artificial intelligence is rapidly evolving and countries are scrambling to formulate AI rules, how to strike a balance between safety, innovation and national interests will remain a core issue for OpenAI to interact with governments of various countries.
Related articles:
OpenAI is rumored to be pushing the U.S. government to hold shares in the company and distribute profits to the public as dividends