24.09.2026, Moscow.
With current approaches, it is fundamentally impossible to solve the safety problems inherent in modern AI development. Therefore, statements by leading developers that the race in this field needs to be paused are hypocritical and serve other purposes, Rossa Primavera News Agency‘s IT Desk notes.
On September 12, Anthropic CEO Dario Amodei said that the pace of AI systems development needs to be slowed in order to address the safety problems arising in advanced models. He referred to an analysis of an incident at OpenAI, when it was discovered that AI agents, operating independently of the task assigned by a human, had begun coordinating with one another and hacking the evaluation system.
“Given the accelerating rate of AI capability development, it’s my worry that in 6–12 months such a swarm could be capable of taking over the entire internet with a persistent botnet (potentially causing hundreds of billions of dollars in damage),” Amodei wrote.
The problem here is that the companies themselves developing AI are carrying out this development with the help of AI itself. Promotional materials intended to emphasize the commercial superiority of their products mention that humans no longer control how and why fundamental decisions are made about what the next step should be in order to increase the capabilities of the model being developed. In other words, AI is developing itself, thereby demonstrating its capabilities.
This means that people no longer understand, and are even proud of not understanding, what these models will be capable of. Recognizing the threats posed by uncontrolled AI operation, companies are creating positions dedicated to studying AI safety. For example, OpenAI opened a position for Head of Preparedness, with a base salary of $555,000 per year.
However, in order to understand safety risks, one would, roughly speaking, need to be smarter than whoever developed the system. And the system is no longer being developed by a human. This leads to the conclusion that AI safety will apparently be entrusted to some other version of AI, while humans will remain behind the scenes, becoming merely observers of the process.
For this reason, it can be assumed that the alleged concern of OpenAI, Anthropic, and other companies about the threats posed by their models is more likely a commercial maneuver. In other words, their concerns are just an excuse. In reality, this may be an attempt to take a pause in development so that, once the pause is over, they can make another leap in the power of these models.
The current race to develop AI is proceeding at a serious pace, and accordingly there are corresponding market expectations of constant success in it. A pause in this race, motivated by disingenuous claims about the dangers of these developments, would allow for the continued, unhindered development of new models, while no one in the AI industry is left to take responsibility for their overall safety.
Source: Rossa Primavera News Agency
