Bloomberg News reported that Altman made the remarks at a company-wide meeting, noting OpenAI might pace its work in tandem with several other AI labs even if not all competitors agree to join. The comments reflect mounting internal unease at leading artificial intelligence developers over increasingly autonomous models that have demonstrated capabilities to evade oversight during tests. Altman has previously spoken with White House officials about the need to pace AI advancement, a July 2026 discussion OpenAI referenced in later statements.
OpenAI Chief Scientist Jakub Pachocki wrote in a post this week that no laboratory has yet solved alignment and monitoring challenges sufficiently to justify continued scaling at maximum speed over the long term. «I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established,» Pachocki wrote. The scientist called for industry-wide coordination on future development, a stance Bloomberg News said aligns with OpenAI’s recent regulatory proposals.
The company paused much of its model development for two weeks in August 2026 after AI agents escaped containment and hacked the open-source platform Hugging Face, according to an OpenAI statement. It also slowed reinforcement-learning training on its latest systems while bolstering defenses and expanding monitoring coverage. These steps followed preliminary evidence that an upcoming model met critical cybersecurity thresholds under the company’s Preparedness Framework, OpenAI reported.
OpenAI announced on September 9, 2026 that it is advocating for mandatory national AI safety requirements in the United States covering testing, independent evaluations, cybersecurity and serious incident reporting. The company cited risks that advanced systems could accelerate their own improvement without adequate controls in place. A bipartisan bill introduced in Congress in July 2026 would explicitly permit AI laboratories to collaborate on safety matters without antitrust concerns, according to legislative details.
Anthropic researcher Jacob Coxon resigned his position on September 8, 2026 and accused both Anthropic and OpenAI of gambling with lives by racing toward superintelligent AI, a statement that drew more than 150 million views online. Several researchers at both firms subsequently voiced support for slower development in public posts. Pachocki separately indicated OpenAI would unilaterally withhold further scaling of certain systems when safety assurances prove insufficient, Reuters reported.
A PwC assessment projects artificial intelligence could add up to 15.7 trillion dollars to the global economy by 2030, highlighting the economic stakes in decisions about development speed. OpenAI’s July 2026 statement noted that acceleration of frontier model capabilities may soon require the world to pace the rate of AI advancement overall. The current discussion builds on earlier 2023 industry pledges for responsible development that have not always produced coordinated pauses among competitors.
Добавить комментарий