Regulating the power of artificial intelligence is, for now, an impossible task while criminal behavior of its models continues to be made public, casting doubt on the future of this technology.
There are no consensus solutions after being one of the star topics of the UN General Assembly and the summit between Donald Trump and Xi Jinping. But many and varied proposals have emerged, some as creative as they are probably useless.
Washington, which does not want to stop AI for fear that China will take advantage of it, has rescued from the cold war with Russia the idea of a hotline, a red phone, to report serious incidents. Beijing has not even taken notice.
Several North American legislators advocate regulations and a panic button that turns off artificial intelligence if it gets out of control. It is also suggested by the Bank of England, which, like the European Central Bank and the US Federal Reserve, has warned of the possible effect on the markets of unleashed AI. But, like viruses and cyberattacks, it would not be easy to shut down a super program that travels and replicates across the Internet.
The main companies in the sector, OpenAI and Anthropic, vehemently defend a regulatory framework that gives confidence to their products and, in addition, clears the financial horizon for their IPOs.
Since Trump does not support them, they look for ways to regulate themselves. Among them a mutual verification of their models that, given the fierce mutual competition, can be as useful in avoiding a Terminator as the chat group on AI opened by the rulers of Canada, France and Norway.