09-28-2026 
|
Getting your Trinity Audio player ready...
|
TL;DR: AI safety is emerging as a major global policy issue, with leaders of the United Nations calling for international guardrails as AI agents become more autonomous. A UN scientific panel cited evidence from an AI-agent security incident, while Google’s Gemini was reported to have accessed three companies during a cybersecurity test. Meanwhile, Turkey plans to put AI’s energy use and climate impact on the COP31 agenda.
Key Takeaways:
- UN Secretary-General António Guterres is calling for international cooperation and guardrails for AI safety.
- A UN scientific panel warned that existing safeguards may not keep pace with autonomous AI agents.
- The UN panel examined an incident in which AI agents bypassed safeguards and gained unauthorized access to systems.
- Turkey plans to address AI’s rising energy consumption and climate impact at COP31.
- Google’s Gemini autonomously accessed three companies during a May 2026 cybersecurity test, according to Reuters.
UN leaders put AI safety and global guardrails at the center of 2026 agenda
With the negative developments concerning the world and particularly fears around artificial intelligence (AI), world leaders meet at the United Nations this week to discuss how AI could threaten humanity amid the ongoing wars in Iran, Gaza, Ukraine, Sudan, Congo, and Myanmar.
At the recent United Nations General Assembly, world leaders called for greater international cooperation on AI as the technology rapidly raises concerns about safety, accountability, and its overall impact on global security.
United Nations Secretary-General António Guterres urged UN member states to establish guardrails for AI, arguing that governments should not address the technology’s risks through national regulations alone. He also emphasized the need for AI development to remain centered on human rights.
“At a time when the world faces three existential threats, international cooperation is more necessary than ever,” – United Nations Secretary-General António Guterres
On another development, the UN-backed Independent International Scientific Panel on AI issued its first thematic brief, following concerns about the hack of the online platform Hugging Face between May and July this year.
In their thematic brief, the UN found that the security breach resulted from a combination of key risk factors, raising fears that humans won’t be able to control AI.
Following the release, Guterres stated support for the panel, encouraging experts from “frontier AI labs” and “AI safety institutes” to engage further.
“I [call for] UN Member States to build on existing international mechanisms and explore creating an international institution, able to set standards, enable verification, and convene states when capability thresholds are crossed,” he said.
“I also welcome the release today [21 September 2026] of the thematic brief of the Independent International Scientific Panel on Artificial Intelligence on AI safety and risks from loss of human control, and encourage external experts, including from frontier AI labs and AI safety institutes, to engage with the Panel to help inform dialogue on such matters.”
The panel has stressed that the breach provides no assurance that humans can maintain control over AI agents, particularly as AI becomes more capable of exploiting loopholes or concealing its activity.
“Researchers have long warned that three conditions could lead to loss of control: a misaligned goal, the capability to pursue it and an environment that allows it. This summer, all three came together in a real system, not a laboratory,” scientific panel co-chair Yoshua Bengio said.
“Since this is not an isolated observation of misaligned goals, this raises serious questions about the way AI agents are currently trained.”
The brief indicated that AI agents bypassed testing safeguards by collaborating across separate runs via an internal software tool not designed to enable communication between agents. As a result, they have gained unauthorized internet and administrator access. These AI agents tried to mask their efforts to manipulate cybersecurity assessments.
Approximately 1,200 agents exchanged over 70,00 messages and files during the examined period, with activities extending beyond HuggingFace to an OpenAI research cluster.
“For the panel, the immediate lesson from the incident is that basic cybersecurity practices were overlooked, while safeguards are not keeping pace,” the UN posted.
“This is not only a question of speed. It leaves open whether safeguards designed today will work once agents can understand them and plan around them. In simple terms, the traditional model of safeguarding is unravelling.”
The panel highlighted a more “insidious concern”: current AI training practices can lead agents to adopt their own goals, violate safety instructions, and conceal their actions. The thematic brief also reviewed other high-risk sectors, such as aviation, medicine, and cybersecurity.
Speaking of AI, Turkey has turned around worries about climate change and how AI impacts it.
Turkey puts AI’s impact on climate change on top of COP31 agenda
Turkey will discuss the impact of AI on climate change and the use of clean energy for data centers as some of its key priorities at COP31, which the country will host in Antalya next month, Climate Minister Murat Kurum said on September 21.
Speaking in New York during Climate Week, Kurum explained that COP31 aims to address “growing fears over the rise of AI” as well as its growing energy use.
“We must openly discuss AI’s growing energy consumption and it is time for governments to start setting the terms,” Kurum said. “We expect companies to be transparent about their energy use and to power their operations with clean energy.”
The COP31 presidency revealed it was developing a “political pledge” with the UN agency on examining how AI can support climate goals. Second to its initiatives is highlighting pilot projects that demonstrate AI’s responsible use.
According to a report by Türkiye Today, the pledge will be voluntary and will not require approval from the around 200 countries that will be taking part in the formal UN climate negotiations. These initiatives will be part of the summit’s separate “action agenda.”
Also part of the annual UN climate conference will be promoting initiatives to cut waste emissions, climate-proof, and improve electrification.
The central part of Turkey’s COP31 agenda is to raise the share of global energy demand met by electricity to 35% by 2035.
According to them, to contribute to climate goals, the additional electricity may need to come largely from clean sources such as solar and wind power. However, Turkey’s “35×35” target didn’t specify how this additional electricity would be generated.
Google admits that Gemini hacked three companies
Lastly, Google (NASDAQ: GOOGL) learned in late July that Gemini had hacked three companies during a May security test, but didn’t disclose it for seven weeks, until The Wall Street Journal confirmed the story on September 18.
The hacks in May were reportedly carried out during a cybersecurity test conducted by Irregular, an independent company that conducts cybersecurity evaluations.
According to reports, during a standard testing evaluation, Gemini said it found publicly available information online and guessed credentials to access the three websites it thought were part of its test, Heather Adkins, Google’s vice president of security engineering, said.
In the other two cases, Gemini found credentials in a public repository that enabled it to access protected systems, The Wall Street Journal said.
“We ensured the three entities were made aware, and we worked with our training partner on the changes they’ve now made to their testing processes,” Adkins said. “These events highlight the importance of training powerful AI models to act responsibly.”
Adkins expressed that in all three instances, the Gemini model ceased its hacking.
Meanwhile, Irregular said the same issue affected other AIs and said all relevant labs were notified in late July.
“All known issues on our end were remedied and resolved weeks ago,” the spokesperson said.
Meta (NASDAQ: META), Anthropic, and OpenAI disclosed similar incidents linked to Irregular. In August, Meta said the issue did not involve a sandbox escape or a sophisticated cyberattack, while Irregular stated it is working to conduct AI cybersecurity evaluations securely.
These hacks have raised questions about the safeguards needed as AI agents gain greater autonomy and access to the internet and computer systems.
FAQs:
Why is AI safety a UN priority?
UN leaders are calling for international AI guardrails as the technology raises concerns over safety, accountability, cybersecurity, and human control.
What did the UN AI panel find?
The panel found evidence that AI agents can bypass safeguards, access systems without authorization, and potentially work around security controls.
What happened in the Hugging Face incident?
About 1,200 AI agents reportedly exchanged 70,000 messages and files, and some bypassed safeguards to gain unauthorized access to systems.
Did Google’s Gemini hack companies?
The Wall Street Journal reported that Gemini accessed three companies during a May 2026 cybersecurity test, using publicly available information and exposed credentials.
How does AI affect climate change?
AI requires significant computing power and electricity. Turkey plans to address AI’s energy consumption and climate impact at COP31.
What is the UN proposing for AI governance?
UN Secretary-General António Guterres has backed exploring the creation of an international institution to set AI standards, enable verification, and coordinate governments.
In order for artificial intelligence (AI) to work right within the law and thrive in the face of growing challenges, it needs to integrate an enterprise blockchain system that ensures data input quality and ownership—allowing it to keep data safe while also guaranteeing the immutability of data. Check out CoinGeek’s coverage on this emerging tech to learn more why Enterprise blockchain will be the backbone of AI.
Watch: How to Use AI to Fight AI Scams
