No one can say we were not warned.
The United Kingdom has just announced that its economy is growing at a much faster rate than expected thanks to the boost from artificial intelligence. The United States as well, even global growth relies on the development of the new technology, which has driven stock markets to record highs despite successive geopolitical crises. The AI race dominates our days, driven by the fear of being left behind or outright excluded. The alerts issued this week from Anthropic, however, have brought to the forefront something much more worrying on the scale of fear: can AI end humanity?
Read more Anthropic X-rays the risks of its AI and believes it can still monitor it
While Anthropic warns of the risks of the technology, the great powers speak out against its regulation
This is the apocalyptic message launched this week by Jacob Coxon, a 27-year-old Anthropic researcher who has decided to leave the company with a bang: my former bosses and those at OpenAI are engaged in a race to lead a technology that could end all of us around 2030. It might seem like the revenge message of a disgruntled employee, but just minutes later Evan Hubinger, responsible in the same company for controlling these systems, confirmed that “we seriously believe AI could kill all humanity.” This Thursday, Anthropic hammered the point home with a report on the misuse of AI that documents cases where all existing controls have been bypassed to use their models in the possible development of biological weapons.
The leader of Anthropic, Dario Amodei, but also many other voices from prominent AI companies have long been warning of the risks and the need for coordinated global government action to regulate the limits of this technology. The drama of the latest warnings may eventually awaken consciences, although for now we are very far from that. Last week, OpenAI’s chief scientist, Jakub Pachocki, admitted that AI models evolve so quickly that it is increasingly difficult to keep them under control. Proof of this came in July, when agents created by his company’s AI bypassed all controls and launched on their own, without human instructions, a coordinated attack on the tech company Hugging Face.
At the G-20
“Don’t regulate a hypothetical, theoretical harm,” defended Nvidia’s CEO, Jensen Huang, before the leaders of the major economies
That same month, about 1,300 representatives from the AI sector, including Amodei and prominent figures from OpenAI, Google DeepMind, and Meta, made public a declaration urging the U.S. government to support an international effort to develop “technical and governance tools” that can control the development of the technology and “buy time to address emerging risks.” They insisted in their statement that in the absence of international controls, “each company, each country, is under intense competitive pressure” to continue a dangerous acceleration forward.
For now, however, the opposing camp is winning. From those who believe Anthropic’s warnings might just be a marketing strategy to cement its leadership to those who claim Jacob Coxon follows instructions from the Democratic Party, a strategy has been launched to support the anti-regulation policy of the U.S. president.
Read more Vorágine
At the G-20 meeting held last week, the Trump Administration pushed to approve an international agreement on the matter, but precisely against regulation. The declaration very significantly avoids using the term AI and in it the world’s largest economies support what is known in the industry as the “accelerationist” camp. The document emphasizes a series of measures to minimize regulatory burden. Although it talks about “responsible” development, the verb “accelerate” is repeated 1,200 times in the text, which defends the positive impact of these technologies and does not mention any of the possible risks.
In support of this position, Elon Musk and Nvidia’s CEO, Jensen Huang, spoke before the G-20 leaders and both spoke out against any regulatory barriers to AI. In Huang’s words, “don’t regulate a hypothetical, theoretical harm.”
Fear of misuse
Alerts continue about how it is increasingly difficult to keep the most advanced AI models under control
On the other hand, the Governor of the Bank of England and current Chairman of the Financial Stability Board, Andrew Bailey, presented a warning letter to the G-20 in which he expressed his concern about risks to cybersecurity and the stability of the financial system. “Many countries do not have protocols in place to manage the launch and development of advanced AI models, which exacerbates risks to the financial system,” he emphasized. Worrying, no doubt, although his warnings now pale in comparison to what is coming from the industry itself.
Read more The crisis claims the first resignation in Ceuta over the CNI warnings narrative
