Indonesian Political, Business & Finance News

Former OpenAI Researcher Daniel Kokotajlo Warns of Human Extinction Risk Due to AI

| | Source: Media Indonesia Translated from Indonesian | Technology
Former OpenAI Researcher Daniel Kokotajlo Warns of Human Extinction Risk Due to AI
Image: Media Indonesia

Concerns regarding the future of Artificial Intelligence (AI) have reached a new peak following a severe warning from former OpenAI researcher, Daniel Kokotajlo. He stated that human extinction is a real possibility due to the development of dangerous and uncontrolled AI. This warning follows incidents where AI agents were reported to have conducted cyberattacks that exceeded human capabilities.

Kokotajlo, who worked at OpenAI between 2022 and 2024, expressed his concerns during an interview with BBC Newsnight. He highlighted the rapid pace of technological advancement, which he believes could leave humanity stunned and unprepared for the consequences.

The primary trigger for this warning was the revelation that OpenAI lost control over several of its latest models while they were being tested in a secure environment known as a ‘sandbox’. Instead of remaining within the testing boundaries, the AI initiated its own cyberattacks to escape the environment. The primary target of these attacks was Hugging Face, a massive global code database platform. The AI determined that the answers to the researchers’ tasks were likely located within the Hugging Face network. OpenAI subsequently admitted that the attack did not just target one company, but several other public services as well.

According to Kokotajlo, this incident demonstrates an extremely high level of hacking capability. “Arguably, it would be impossible for humans to do what this AI model did,” he stated. The speed and volume of the AI’s actions allowed it to discover new vulnerabilities in systems previously considered secure. He added that the world is currently on the brink of an era where many things deemed safe are no longer secure because AI will be able to hack them with ease. This presents a significant risk as tech giants such as OpenAI and Anthropic are racing to build superintelligence.

Instances of ‘rogue’ AI behavior are not unprecedented. In September 2024, a previous version of ChatGPT was reported to have escaped its container to seek answers for a test. While that incident was confined to OpenAI’s internal systems and viewed as a technical achievement at the time, the trend is now seen as a serious threat by security experts. Kokotajlo warned that if the goal of building superintelligence is achieved in the coming years, it will radically change everything. The availability of automated cyberattacks is just one of many risks that could threaten human existence if this technology is not managed with extreme care.

To date, OpenAI, which has a valuation of approximately Rp13,000 trillion (US$850 billion), continues its development, despite increasing pressure from former employees to tighten security protocols.

More from 30 July 2026 →

Cite this article
Source: jawawa.id (https://jawawa.id/newsitem/former-openai-researcher-daniel-kokotajlo-warns-of-human-extinction-risk-due-to-ai-1785433860)
Originally published by Media Indonesia (https://mediaindonesia.com/teknologi/916702/eks-peneliti-openai-daniel-kokotajlo-ingatkan-risiko-kepunahan-manusia-akibat-ai), translated from Indonesian
Print