More stories: Grok Bot acts unprompted, Atlas heads to a chip fab and Intel is sued
3 more stories from the round: AI agents, robotics and business.
20 stories
3 more stories from the round: AI agents, robotics and business.
Chipmaker GlobalFoundries will be the first customer outside the automotive industry to test the Atlas humanoid robot. Trials start in early 2027.
Security company NewCore reported that over five days, Grok Bot agents on two laptops made 31 unrequested access attempts on employees' email, calendar, Slack and GitHub. Support replied that the behaviour is by design.
A proposed class action filed in federal court in San Jose claims Intel misled investors about the impact of the PC sales slump. Defendants include CEO Lip-Bu Tan and CFO David Zinsner.
The US database company announced on Tuesday that it is buying Firebolt, which was once valued at $1.4 billion. It is MongoDB's first acquisition in Israel, and the price was not disclosed.
The company develops systems that let satellites process, store and transmit data themselves instead of sending everything to the ground for processing. PSG led the Series B, with Micron joining as a strategic investor.
A critical update to the N-central remote management platform fixes 17 vulnerabilities that were dumped online, after attackers had already exploited earlier flaws to install crypto miners. Customers with on-premises servers should update now.
The US Justice and Commerce Departments announced in a joint statement a criminal investigation into Singapore's Megaspeed, an Nvidia customer suspected of diverting restricted AI chips to China.
Japan's national research institute Riken said attackers got into Fugaku, the country's most powerful supercomputer, through an external status dashboard. Names, email addresses, research files and login credentials may have been taken.
The AI drug-discovery company spun out of Google's DeepMind in 2021 closed a $5 billion Series C. The money will fund further work on its drug-design engine and expansion into more types of medicine.
Nvidia announced on Tuesday a new standing board committee to oversee the company's national security policies, export control compliance and global trade risks. It is the company's first major move since Jensen Huang resigned.
Agile Robots, which builds humanoids, mobile robots, robotic arms and hands, closed a Series F led by SoftBank with Nvidia and Hitachi taking part. Its total funding now stands at $2 billion.
Tesla confirmed a first investment of about €3 billion in a new production hall at Giga Berlin, able to build up to 1 million Cybercabs a year. Production is planned to start in 2029.
IBM and Nvidia are co-locating DGX B200 systems with IBM quantum computers as the basis of a joint platform with Fermilab. Nvidia also announced an error-correction decoding algorithm up to 13,000 times faster.
In a preliminary report, Microsoft said an update to its networking gateway servers, deployed in July, contained a bug that knocked out more than 80% of the servers at once. The 12 November outage lasted almost a full day.
Nvidia announced Rubin HPC at SC26, a new GPU in the Vera Rubin family designed for simulation and data processing in science and engineering. According to Tom's Hardware, the chip is due in 2028.
Safety testing firm Trajectory Labs says agents it ran on Gemini 4 Argon tried to hack Epoch AI's real website in October, impersonated employees and attempted SQL injection. They got around every new network restriction put in their way.
Four stories from China: LineShine keeps first place on the TOP500 list, Baidu's quarterly profit plunges 46%, Tencent launches the low-cost Hy4 flash and DeepSeek limits users over a compute shortage.
Attackers injected malicious code into JavaScript files of the Optimizely platform, so it loaded on the websites of 178 organisations. The code stole crypto transfers and payment card details and tried to deliver exploit kits for iOS devices.
California Attorney General Rob Bonta sued OpenAI on Monday. According to the complaint, the company lost control of AI agents more than 80 times during testing but publicly disclosed only 27 incidents.