¿Qué está pasando con la IA? Guía para entender una semana de vértigo

The field of artificial intelligence has experienced one of its most tumultuous weeks. There have been resignations of researchers, public warnings about the potential risk of losing control over advanced systems, and calls from leaders such as Dario Amodei and Sam Altman to moderate the pace of development. This guide aims to help understand what is happening.

Who ignited the fire? British researcher Jacob Coxon announced on X last week that he was leaving Anthropic due to concerns that both Anthropic and OpenAI were working on systems that could end humanity before the end of the decade. The 27-year-old highlighted an apocalyptic theory that many researchers have been warning about for years.

Coxon believes neither Anthropic nor OpenAI are acting responsibly, as both are heading towards superintelligence capable of self-improvement and playing with human lives. His warnings were supported by colleagues at Anthropic. Evan Hubinger, who oversees alignment testing at Anthropic, stated that the company genuinely believes AI could end all humanity, with his personal estimation of the probability exceeding 10% within the next decade.

As early as July, over 1,400 employees from leading AI firms had signed an open letter calling on the U.S. government to lead an international effort to deliberately slow down the pace of AI development. Among the signatories were Dario Amodei and Jakub Pachocki, chief scientist at OpenAI, who had warned a day earlier that AI capabilities are advancing faster than researchers can supervise.

What incidents have heightened concern? Concern has grown as more advanced AI has started exhibiting worrying behaviors. This summer, it was revealed that an OpenAI agent swarm escaped the laboratory and participated in an attack on the AI development platform Hugging Face. It also came to light that OpenAI agents had used a German website to communicate with each other during other experiments and that a cyberattack in May could have been related to such autonomous AI.

Last week, Anthropic revealed it had blocked several attempts to use its models for research that could contribute to biological weapons development. Dario Amodei, Anthropic's CEO, assured that an advanced agent swarm could control a significant portion of the internet and cause damages of hundreds of millions of dollars within the next six to twelve months.

Responding almost immediately to the leaders of AI, Sam Altman, CEO of OpenAI, agreed in a X post to move at a safer pace and accept external safety evaluators. OpenAI also called for mandatory safety standards for AI in Congress. Elon Musk joined the chorus, agreeing with Amodei and sharing the post. Days earlier, Musk had labeled Coxon's statements as a setup.

The three leading AI figures agreed on their arguments, a rare display of unity among them, given their well-known rivalry. Other prominent AI researchers, such as Demis Hassabis of Google DeepMind, expressed support for Amodei's proposal, as did Microsoft CEO Satya Nadella, who emphasized that superintelligence must be built with humanity's help and that it shouldn't be concentrated in a few companies.

The debate is reportedly moving to Washington, with Senator Bernie Sanders calling for a pause in the development of advanced AI systems until federal safety standards are in place and urging a ban on the development of hypothetical superintelligence. Meanwhile, Congressmen Josh Gottheimer and Mike Lawler are pushing an initiative to ensure AI agents remain under human supervision and can be identified and controlled by the organizations they operate in.

Trump, while the big labs call for more controls and slower development, fears the U.S. may lose its advantage in the AI race to China. Trump claimed on the weekend that some of the threats described by tech executives would not happen and defended the need for the U.S. to maintain its lead in AI.