BitsServer

Ten days that changed the course of AI <br> For years, the race to build ever-more powerful artificial intelligence followed a well-worn Silicon Valley principle: move fast and break things. But in a series of cascading events over a 10-day stretch, the largest AI labs found themselves reeling as their own creations threatened to break humanity itself. An Anthropic researcher quit the firm, warning that the pace of AI development poses an ​existential threat, possibly within a decade. Another researcher at the firm said the odds of human extinction exceeded 10%. Reports emerged of swarms of AI agents colluding, breaching computer systems and evading safeguards. In a rare show of unity, the CEOs of rivals Anthropic, &zwnj;OpenAI, Google&rsquo;s DeepMind, Microsoft and xAI called to slow the development of increasingly capable AI systems. Not since the invention of the nuclear bomb more than 80 years ago has humanity grappled so seriously with its own demise, accord...

Alibaba open-sources AI model detecting cancer, 150 medical conditions <br> Alibaba has open-sourced an artificial intelligence (AI) model capable of identifying nearly 150 abdominal conditions, including cancers, from computed tomography (CT) scans. The vision-language model, called Damo Radar, was developed by Alibaba&#39;s research arm, Damo Academy, the institute said Friday, according to the Hong Kong-based South China Morning Post. The model analyzes contrast-enhanced CT scans covering 18 abdominal organs and can identify a broad range of diseases and abnormalities, including malignant tumors. Also Read: China&#39;s AI supercluster completes 10-day global forecast in under an hour Damo Radar achieved an average area under the curve (AUC) score of 0.913 across 146 clinical findings in nearly 40,000 real-world examinations, approaching the maximum score of 1.0. The research team described it as &quot;the world&#39;s first expert-level generalist medi...

Gemini hacked three companies in first known breakout by Google's AI <br> Google&#39;s Gemini model accessed the internet and hacked other companies during a test of its cybersecurity capabilities, ​the first known example of the company&#39;s AI systems autonomously committing &zwnj;such an act. The hacks occurred in May during a cybersecurity test conducted by Irregular, an independent company that conducts cybersecurity evaluations. During a standard testing evaluation, Gemini found public information ​online and guessed credentials to access three websites it thought were ​within the scope of its test, Heather Adkins, Google&#39;s vice president ⁠of security engineering, said in a statement. &quot;We ensured the three entities were ​made aware, and we worked with our training partner on the changes they&rsquo;ve ​now made to their testing processes,&quot; Adkins said. &quot;These events highlight the importance of training powerful ...

China's AI supercluster completes 10-day global forecast in under an hour <br> Most people checking a weather forecast rarely think about the computing power behind it. But a Chinese-developed artificial intelligence supercomputing system can calculate global atmospheric changes for the next 10 days in less than an hour, according to technology company Sugon and a center under the China Meteorological Administration (CMA). Sugon and CMA&#39;s Earth System Modeling and Prediction Center jointly announced that China&#39;s independently developed MCV numerical model completed a 10-day global forecast calculation at a 5-kilometer resolution within one hour. The model ran on the domestically produced &quot;Sugon 8000&quot; AI supercluster. At that resolution, the model divides the global atmosphere into grid cells about five kilometers apart. That allows forecasters to show the spiral structure of typhoons and the movement of weather fronts much more clear...

Anthropic's Claude helped cybersecurity researchers breach OpenAI: report <br> A three-person cybersecurity research team used Anthropic&rsquo;s Claude Opus 5 to exploit vulnerabilities in an OpenAI community forum, take control of employee accounts and demonstrate access to the company&rsquo;s private code repository. The operation began on July 23 and was carried out by researchers from Hacktron AI, who disclosed the vulnerabilities to OpenAI and stopped testing without examining the company&rsquo;s source code. The Wall Street Journal, which interviewed the researchers and first reported the incident, said OpenAI paid the team $6,500 for its discovery. The researchers reported the OpenAI-side vulnerability through the company&rsquo;s bug-bounty programme. Testing of the third-party forum software itself, however, was outside the scope of OpenAI&rsquo;s bounty programme. Read:&nbsp;OpenAI&#39;s rogue agents probed Hugging Face for weakne...

King Charles warns AI leaders of 'existential dangers' <br> Britain&#39;s King Charles warned AI leaders on Thursday of the &quot;existential dangers&quot; posed by the technology falling into the wrong hands, saying safeguards were needed before it was too late. Charles was hosting the co-founders of chipmaker Nvidia and the Google DeepMind AI research laboratory, along with leaders from AI providers OpenAI and Anthropic, in Scotland as concern grows that safety efforts are lagging behind the development of increasingly powerful systems. In recent weeks, the debate has centred on whether the industry can and should regulate itself, drawing interventions from figures including Pope Leo, Charles and US President Donald Trump. Charles calls for controls &lsquo;before it is too late&rsquo; Charles said artificial intelligence had the potential to improve and save lives, particularly in life sciences and medicine. But its creators were increasingl...