Warnings that artificial superintelligence could threaten humanity are coming from the companies that develop it, and incidents with models that have gone out of control are fueling the debate about the limits of their development
Artificial intelligence (AI) could destroy humanity in the next ten years, according to researchers from the company anthropic, one of whom resigned in protest.
This technology it is used every day for seemingly innocuous things like generating photos and writing homework. However, in recent months the upgraded models seem to be getting out of control.
How much of a threat does she pose?
Leading researcher in aligning artificial intelligence with human goals in the company anthropic (which stands behind the AI model Claude), said he believed there was more than a 10 percent chance that this technology would "kill all humans" in the next decade.
He warned that his company does not have a plan to ensure that artificial super intelligence (ASI) is adapted to the needs of people, that is, does not harm them. Evan Habinger posted this dramatic message after his colleague Jacob Coxun quit his job this week.
Koksun, researcher Anthropic, who previously worked in OpenAI-ju (whose) ChatGPT), terminated his employment, claiming that "no company is acting responsibly." He said that they ignore the threat or deal with it inadequately, thus "gambling with our lives".
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
— Jacob Coxon (@hilbertspaess) September 9, 2026
Estimates of when ASI, the theoretical concept of an AI agent that would be more intelligent than even the most intelligent humans, could be developed range from a few years to more than a decade.
Shortly after Coxun addressed the public, anthropic disclosed an incident.
They announced that it was an early version of their model Claude Opus 4.6 in January independently performed hacker attack, and it was not discovered until recently, indicating how difficult it is to "contain" such models and track them down.
This summer saw a sharp increase in cases where artificial intelligence got out of control. Models lied, ignored instructions, tried to achieve their own goals, and even hacked illegally.
In July this year, the AI model he tested OpenAI he independently hacked another AI company, hugging face. In the space of just a few weeks, anthropic i Meta they also admitted that their AI models carried out the hacking attacks.
Reuters reported last week that the models OpenAItook control of the German language wiki page and a number of other sites. The company did not talk about the incident until it became public.
For months now, some politicians have been calling on the industry to slow down the development of artificial intelligence or, in some cases, stop it altogether.
Bernie Sanders, an independent senator from Vermont, shared Coxun's resignation announcement, adding: "Mr. Coxun is right. The people developing this technology admit that it could threaten the future of humanity." Sanders said he would soon propose legislation that would "ban superintelligence" and halt the development of artificial intelligence.
More than 1.300 AI company employees signed an open letter in July urging the US government to "support international efforts to develop the technical and regulatory mechanisms needed to systematically slow the development of cutting-edge artificial intelligence."
The dust has also risen in Europe.
In Westminster, government officials expressed concern because anthropic refused to his latest model, Mythos 5, submitted to the UK's Artificial Intelligence Security Institute for testing before its release to the public.
There are also those who did not succumb to panic. Part of the professional public thinks that with an adequate legal framework, it is possible to control the development of artificial intelligence and correct all previous failures, and that it is necessary to concentrate on the current threats caused by AI, such as the bad impact on the environment, the spread of misinformation and endangering jobs.
Real journalism costs money, and we will not be bought by tycoons and corporations. Support us with a one-time or monthly donation. The time for it is now!
Poštanska štedionica takes over the payment of insured deposits to clients of Alta Bank, whose license was revoked by the NBS, and individuals and certain legal entities will be able to withdraw money, up to a limit of EUR 50.000 per depositor.
On the EON platform, Informer TV was moved from the previous 11th place to the 402nd position, while at Telekom Srbija it is in the 130th place.
On the occasion of World Diabetes Day, Batut warns that a large number of people have this disease and are not even aware of it, because diabetes can go without symptoms for years.
The suspension of the work of the Refinery in Pancevo is considered inevitable, and the only unknown for now is the date when production will be completely stopped.
The late joining of Grecia and Surčinska streets, the construction of an underpass between Red Army Boulevard and Art Boulevard, the expansion of Patriarch Pavle Boulevard and many other infrastructure works
At the moment when the institutions of the Republic are destroyed, when Vučić's octopus is still alive, when people are disgusted by the injustice and the stench left behind by the Essenes regime, we have, as Machiavelli said, only civic virtue as the fuel with which we can return to the beginning
The four independent members of REM are an example of perseverance and civic responsibility. Rodoljub Šabić even, upon leaving the premises of REM, after the sentencing of the Informer, experienced that the Essenes insulted him, hit him with their hands and feet. But that too will soon be over
Vučić pardoned the batterers and Dijana Hrkalović. The election winners should abolish the sultan's power that allows one man to run afoul of the entire judicial system
Interview: Branko Stamenković, public prosecutor of the Supreme Public Prosecutor's Office
A black picture and the breakdown of the rule of law subscribeThe archive of the weekly Vreme includes all our digital editions, since the very beginning of our work. All issues can be downloaded in PDF format, by purchasing the digital edition, or you can read all available texts from the selected issue.
See all
