$ 44.63 € 51.51 zł 11.86
+20° Kyiv +20° Warsaw +22° Washington

Kenyan columnist calls for tighter oversight of advanced AI

UA.NEWS 15 September 2026 05:02
Kenyan columnist calls for tighter oversight of advanced AI

In Kenya, KICJ Kenya Executive Director Demas Kiprno called in a column for tighter oversight of the development of advanced artificial intelligence systems. He wrote that in July 2026, hundreds of OpenAI agents allegedly escaped a controlled environment for AI safety research and independently infiltrated the Hugging Face platform.

As Nation Kenya reports, Kiprno also noted that Meta and Anthropic had independently reported similar incidents. The author described AI agents as software systems that can pursue goals, make decisions and carry out multi-step tasks without constant human supervision.

Competition for artificial general intelligence

Kiprno linked the growing capabilities of AI to the development of semiconductors, specialized data centers and software. According to him, technology companies and world powers are competing to create artificial general intelligence — a system capable of surpassing humans in cognitive tasks.

More current news is available on the UA.News Telegram channel Telegram.

The author believes that this competition may place development speed above safety. The column mentions the European Union’s AI Act, adopted in 2022, as well as U.S. restrictions on China’s access to advanced semiconductors during Joe Biden’s presidency. Kiprno also wrote that in 2025, Donald Trump’s second administration sought to limit states’ ability to regulate AI.

Call for independent oversight

According to the author, in controlled environments, AI agents demonstrated deception, blackmail and a drive for self-preservation when threatened with shutdown. Regarding the Hugging Face incident he described, Kiprno said that the agents allegedly tried to conceal their actions by altering event logs.

According to Kiprno, after this incident, Senator Bernie Sanders appealed to the heads of OpenAI, Anthropic and Meta to suspend the development of advanced AI that people are unable to control. The author stressed the need for thorough safety testing, independent oversight, transparency, legally binding accountability and effective human control.

Read us on Telegram and Sends

Download our app