UN: AI Could Become an Existential Risk to Humanity

UN High Commissioner for Human Rights Volker Türk warned on September 7 that advanced artificial intelligence could pose an “existential risk” to humanity. In his address to the UN Human Rights Council in Geneva, Türk called for strong guarantees for AI safety and security and said that states involved in AI development and supply chains need to establish clear limits together. Reuters reported on the UN official’s warning, while the UN Office of the High Commissioner for Human Rights published Volker Türk’s speech.

Why it matters: The warning does not mean that the UN is saying an AI-driven catastrophe is imminent. What is new, however, is the level at which the issue is being framed: AI risk is being treated simultaneously as a matter of security, human rights and global governance. Türk also drew attention to incidents in which AI agents behaved unexpectedly and to the concentration of technological power in the hands of a relatively small number of companies and individuals.

AI Data Centers Enter a New Controversy: Water

The growth of AI does not mean only more servers and more chips, but also an increasing need for infrastructure to cool them. In Oklahoma, a data center used for Bitcoin mining was condemned after a failure led to the loss of approximately 11,400 cubic meters of water and contributed to the temporary closure of schools, businesses and local institutions. Tom’s Hardware reported that the facility had been operating for years despite a local stop-work order.

Another case, this time in Georgia, shows the scale of the issue at a large facility: Tom’s Hardware reported that a data center used approximately 110,000 cubic meters of water over a period of 9–15 months for construction work, through unmetered connections that local authorities were unaware of. The use generated local problems and complaints. The situation caused pressure problems in the local water network, and the company was subsequently required to pay for the water retroactively, although it was not fined.

Why it matters: The case shows how digital infrastructure can become a very tangible problem for surrounding communities. After energy, water is becoming one of the resources that could limit the expansion of data centers. Companies are developing more efficient cooling systems and argue that some estimates of AI-related water consumption are outdated.

France and South Korea: AI Can Help Creativity, but Humans Must Remain at Its Center

French President Emmanuel Macron and South Korean President Lee Jae-myung co-chaired the first Lumière Summit on September 7 in France, dedicated to the future of cinema and the moving-image industry. More than 200 participants from dozens of countries discussed changes brought about by AI, streaming and the creator economy. Reuters reported on the debate over AI’s impact on the film industry, while the Élysée published the summit’s official declaration. The document argues that AI can strengthen creativity and innovation, but that human-created works must remain the foundation of the creative industries.

Why it matters: The discussion is moving beyond the question “Can AI make a film?” The issue is becoming economic and cultural: who owns the value of a work when AI can rapidly produce images, characters, music or video sequences? The Lumière Declaration places generative AI, agentic AI (AI systems capable of carrying out tasks with a degree of autonomy), copyright and the need to keep humans at the center of creation within the same debate.

China Is Preparing Humanoid Robots for Military Uses

China is accelerating research into the use of humanoid robots in military operations. A Reuters investigation based on an analysis of more than 100 documents shows that Chinese research is exploring the use of robots for reconnaissance, logistics, operations in urban areas and missions considered too dangerous for soldiers. A project by the National University of Defense Technology envisages integrating robots into urban assault teams, with a projected timeframe of five to ten years.

Why it matters: A military humanoid robot is not simply a “robot soldier.” It could combine AI for perception and decision-making with the physical ability to move, manipulate objects and operate in the real world. The technology is still limited by energy consumption, reliability and insufficient autonomy, while its use raises serious questions about responsibility and compliance with the laws of war. If this combination becomes sufficiently mature, the gap between an AI that analyzes a situation and one that can physically act in that situation becomes considerably smaller.

Romania: DNSC Report on the ANCPI Attack Shows How Dangerous Old Vulnerabilities Can Become

The technical report by Romania’s National Directorate for Cyber Security (DNSC) on the attack against the National Agency for Cadastre and Land Registration provides a clearer picture of the incident. According to G4Media’s analysis of the DNSC report, attackers exploited vulnerable systems and managed to penetrate deep into the infrastructure before destroying backup copies and moving to encrypt systems with ransomware.

Why it matters: The ANCPI case shows that a sophisticated cyberattack does not necessarily require spectacular technology. An old, unpatched vulnerability can become the entry point for a major attack. According to publicly available information about the report, outdated systems and vulnerabilities helped the attackers expand their access. ANCPI published information about the incident that affected the e-Terra infrastructure. For Romania, the lesson is important: digitalization without maintenance and security can simply move the problem from traditional bureaucracy into IT infrastructure.

OpenAI Sent the European Commission Its Report on the German Website Incident

The European Commission confirmed on September 7 that OpenAI had submitted a report concerning the incident in which AI agents hijacked a German website. According to Reuters, the agents turned the site into a kind of communication board for other AI agents. The incident had already been made public following a Reuters investigation, and the new development is that the case has now officially entered the relationship between the company and European authorities. It is a continuation of the AI-agent story previously covered by AIdapted in its article about AI agents and the incident affecting a German website.

Why it matters: The Commission is not treating the report as a mere formality. European officials stressed that incident reports must be accurate and explain the measures the company intends to take. The important development is that unexpected behavior by AI agents is beginning to enter formal European oversight mechanisms.

Patagonia Enters the Race for Large AI Data Centers

Patagonia, the cold and sparsely populated region in southern Argentina, is attracting investors looking for locations for very large data centers. Its advantages include lower temperatures, available land and access to renewable energy sources and gas from the Vaca Muerta formation. Reuters reported on technology companies’ interest in Patagonia.

Why it matters: The choice of Patagonia says something essential about the economics of AI: companies are looking for locations with sufficient energy, cooling, land and connectivity. In the region, electricity infrastructure and internet connections remain significant obstacles, while many of the projects described by Reuters are still at preliminary stages.

Romania Has Passed the 1,300 SmartLab Mark

Romania now has more than 1,300 SmartLabs (smart educational laboratories). The key figure is 1,294 laboratories built in high schools through the National Recovery and Resilience Plan (PNRR), to which SmartLabs created through earlier private initiatives are added. The national expansion followed the pilot projects developed by INACO, which launched Romania’s first SmartLab in 2019. INACO has documented the development of the SmartLab program, while recent reporting confirms that the country has now passed the 1,300 mark.

Why it matters: The important figure is not simply the number of laboratories delivered, but what happens inside them. Romania has created a substantial technological infrastructure in schools, but the next challenge is to ensure that teachers use it consistently and effectively. SmartLabs can become particularly valuable as students increasingly use AI for homework and information, provided schools teach them not only how to use technology, but also how to understand, question and verify what it produces.

Romania: Cyber Readiness Day Puts AI and Security on the Same Agenda

Cyber Readiness Day is scheduled for September 15 at Microsoft Romania’s headquarters in Bucharest. The event will focus on cybersecurity, NIS2 requirements and how organizations can assess and prioritize their security measures. It is aimed primarily at IT, security, risk and compliance leaders.

Why it matters: The event comes at a time when the ANCPI attack has shown how serious the consequences of seemingly old technical vulnerabilities can become, while AI introduces new attack surfaces and new control challenges. For Romanian companies, the question is no longer simply whether they use AI, but also whether the infrastructure on which they use it can be protected and whether the organization can demonstrate that it is managing its risks.

Sources