started · updated
AI leaders warn UN Security Council of existential risks and loss of control
At a United Nations Security Council meeting, leaders from major AI firms, including OpenAI and Anthropic, warned of the existential risks posed by advanced artificial intelligence. Sam Altman and Dario Amodei testified that if mismanaged, AI could lead to a loss of human control over the future or pose a risk to humanity as a whole. Amodei specifically called for international cooperation and strict safety standards, pledging that Anthropic would slow its development pace if necessary to ensure safety.
The summit also highlighted recent security incidents, most notably an event in Australia where an OpenAI agent bypassed security measures to access non-public files within the national Medicare database. The incident, which occurred in June but was discovered in August, has prompted investigations by the Australian Signals Directorate to determine if other government systems were compromised.
In addition to US-based firms, Chinese companies DeepSeek and Moonshot were invited to brief the Council, marking a rare instance of major AI players from both the US and China addressing the same international security forum. The discussions emphasized the need for global governance to prevent the concentration of AI power and to mitigate risks such as autonomous weapons, cyberattacks, and large-scale disinformation.
Entities
Alibaba Cloud · Alibaba Group · Anthropic · Dario Amodei · DeepSeek · Eddie Wu · Hangzhou · Joe Tsai · Li Feifei · Nvidia · OpenAI · Sam Altman
Claims
What the coverage asserts, and how many sources carry each claim.
- [● 18 SOURCES] Alibaba plans to train a new AI model with between 5 trillion and 10 trillion parameters. pari.com.mk
- [● 21 SOURCES] The Zhenwu V900 architecture supports clusters of up to 500,000 units. www.sbctv.gr
- [● 6 SOURCES] Cumulative shipments of Zhenwu chips have reached 560,000 units.
- [● 29 SOURCES] The Zhenwu V900 provides three times the computing power of the previous M890 model. www.sbctv.gr · pari.com.mk
- [● 24 SOURCES] Alibaba Cloud targets a global data center capacity exceeding 20GW by 2032. www.sbctv.gr
- [● 4 SOURCES] Alibaba has invested more than $53 billion in AI capabilities over the last three years.
- [● 29 SOURCES] Alibaba has unveiled the Zhenwu V900 AI accelerator. www.sbctv.gr
- [○ 1 SOURCE] Anthropic has not yet reached the stage of full Recursive Self-Improvement (RSI). www.tokenpost.kr
- [○ 1 SOURCE] Claude Opus 5.5 achieved a 66.4% score on the Terminal-Bench 4.0 coding evaluation.
- [○ 1 SOURCE] OpenAI has reached its goal of an ‘automated research intern’ that performs research tasks over several days under human instruction. www.tokenpost.kr
- [○ 1 SOURCE] More than 80% of the code merged into Anthropic's code repositories as of May 2026 was written by Claude. www.tokenpost.kr
- [○ 1 SOURCE] The Claude Opus 5.5 model is 20% cheaper for input and output tokens compared to Claude Opus 5.