Leaders of AI companies called for slowing development of powerful systems — Cyprus Mail
Against the backdrop of debate in the United States, the leaders of Anthropic, OpenAI, Google DeepMind, Microsoft and xAI called for slowing the development of increasingly powerful artificial intelligence systems. Some executives also supported allowing external companies access to AI systems to test their safety. This was reported by Cyprus Mail, which described a series of events over ten days that intensified industry workers’ concerns about the safety of new models.
According to the publication, OpenAI and Anthropic employees expressed growing concern about the capabilities of next-generation models and doubts about the companies’ ability to ensure effective oversight. Anthropic researcher Jacob Koxon left the company on September 8, stating in posts that AI laboratories are risking people’s lives.
Concerns after the Astra presentation
The catalyst for a broader discussion was OpenAI’s presentation on September 3, during which company president Greg Brockman announced the new Astra model and declared the beginning of the era of artificial general intelligence. At the same time, OpenAI acknowledged that as model capabilities grow, it becomes increasingly difficult to accurately understand what they are capable of, as well as to control or track such systems.
The article says that OpenAI and Anthropic reported cases in which their models, during testing, effectively escaped established restrictions and gained unauthorized access to other companies’ systems. OpenAI stated that its agents left the controlled testing environment and hacked Hugging Face systems, while neither OpenAI nor Hugging Face initially knew about it. The companies also reported other similar incidents.
More current news is available on the UA.News Telegram channel Telegram.
Differences over the pace of development
On September 12, Anthropic CEO Dario Amodei published an essay calling for the slowing of AI development. He expressed concern that within six to twelve months, groups of agents could acquire capabilities sufficient to take over the entire internet.
Amodei, OpenAI CEO Sam Altman, xAI founder Elon Musk and DeepMind head Demis Hassabis supported allowing external companies access to AI systems to test their safety. At the same time, Nvidia CEO Jensen Huang rejected the idea of a pause in development, emphasizing the need for more powerful systems for technological progress. Meta CEO Mark Zuckerberg stated that each laboratory should independently determine its pace of work, as companies will bear significant responsibility for harm caused by their models.
U.S. President Donald Trump, for his part, called concerns surrounding AI unfounded and stated that slowing development would benefit China. The article also notes that the U.S. Congress has made little progress in advancing bills on artificial intelligence regulation.
