Researchers from universities in France and Italy discovered that when people rely on AI advice, their tendency to admit uncertainty drastically decreases—from 44% down to just 3%. At the same time, their accuracy in making correct decisions dropped from 27% to 9%. Interestingly, their confidence surged from 30% to 76%, indicating that people felt more sure despite performing worse. As one researcher noted, people’s accuracy fell to one third while their confidence doubled.
In the latest Equity episode, we discuss the potential effects of Apple's lawsuit on OpenAI's emerging hardware initiatives and its plans to go public.
Jensen Huang concluded his Tokyo visit by securing significant partnerships across Japan's diverse technology landscape, highlighting a comprehensive engagement with the country's tech sector.
As the FIFA World Cup 2026 culminates with Spain facing Argentina for the championship, the impact of the event extends far beyond the matches. With over 75% of the global population engaging in some way, this tournament served as a major platform for technological advancements. Key innovations include AI systems that analyzed team performance and helped fans navigate host cities, advanced ticketing challenges and resale disputes, cutting-edge security and anti-drone technologies, as well as creative design and marketing campaigns showcasing cultural heritage and engaging global audiences. These technologies and approaches are set to influence not only sports but also tourism, public safety, marketing, and event management well into the future.
Over 12 million compromised streaming accounts linked to World Cup broadcasts are actively traded on the dark web, with potential black-market revenue nearing $220 million. These figures come from the Satori Threat Intelligence team at HUMAN Security, which monitored accounts across 10 streaming platforms showing tournament matches. Notably, on June 27—the last day of the group stage—cybercriminals released 802,000 stolen accounts in one day, highlighting the significant scale of this illicit activity.
Raidium, a startup operating out of Paris and Silicon Valley, has introduced its AI-driven imaging platform, Raidium Read, at Moffitt Cancer Center in the US. This innovative tool has replaced the center's previous radiomics software and is now being used in clinical trials and research. The company is anticipating FDA 510(k) clearance soon, aiming to expand its application further in medical imaging.
At the World AI Conference in Shanghai, Alibaba's chip design division, T-Head, announced the open-source release of SAIL, the comprehensive software stack for its Zhenwu AI chips. This initiative aims to break developers' reliance on Nvidia's CUDA platform by providing an adaptable alternative that supports mainstream AI chip programming.
Kevin O’Leary, known for his role on "Shark Tank," recently argued that AI data centers consume significantly less water than golf courses in the U.S. This statement comes amid controversy over his large Stratos data center project in Utah, which has faced local protests and prompted a gubernatorial executive order. While his comparison is accurate based on current data, the reality of water usage in data centers versus golf courses is complex and involves various factors beyond simple consumption figures.
Nebius has successfully obtained $775 million in its inaugural secured debt facility by leveraging its deployed GPU infrastructure and contracted cash flows from a top-tier investment-grade client. This financing arrangement, which matures on October 31, 2030, carries an interest rate tied to SOFR plus 2.50%, currently around 6.8%. Alongside the customer contract cash flows, the facility fully covers the debt, providing strong financial backing for Nebius. The company still has an additional $40 billion in contracts available to securitize in the future.
The Trump administration has taken over the responsibility of deciding which companies and organizations can access advanced AI models developed by Anthropic and OpenAI, according to sources cited by CNBC. Previously, these AI labs had the autonomy to manage who could use their frontier models. Anthropic managed access to its Mythos cybersecurity model through Project Glasswing, while OpenAI controlled its own distribution independently.
