Samuel Thompson

AI stocks declined, triggering a tech sell-off, after a report from AI firm NANDA (MIT Media Lab) highlighted the low success rate of generative AI in commercial deployments. The report found only 5% of pilot projects deliver financial returns, citing issues with contextual awareness and learning capabilities. Successful applications are more common in back-office workflows. While individual productivity with AI is high, institutional-level benefits are lacking, raising questions about the technology’s practical effectiveness and potential hype. The report suggests strategic partnerships with AI vendors, a recommendation aligned with NANDA’s business interests.

At Security Summit 2025, Google Cloud outlined its AI-powered vision for proactive security. Key announcements include AI Protection enhancements within Security Command Center for AI agent discovery and threat protection (like prompt injection). The vision extends to an “agentic SOC” with AI agents assisting in threat management and alert investigation. Google Security Operations receives AI-driven capabilities (Gemini AI), unified dashboards, and upgrades to the Trusted Cloud, focusing on compliance, smarter access control, and expanded data/network security for AI. The goal: secure innovation through AI-embedded security.

Nvidia is reportedly developing a new AI chip, the B30A, for China, aiming to outperform its current H20 offering while complying with US export controls. Utilizing its Blackwell architecture with a single-die design, the B30A is expected to have roughly half the performance of the B300. Nvidia is also purportedly creating an RTX6000D chip for AI inference in China. These moves reflect Nvidia’s strategy to maintain a presence in the critical Chinese market amid US-China trade tensions and regulatory scrutiny. Samples of the B30A may be sent next month.

A UK government report warns of a “once-in-20-years opportunity” to build a world-class AI chip design industry. Failure to invest risks the UK becoming a consumer, not a creator, of this critical technology. The report emphasizes that focusing on design, not manufacturing, is crucial, requiring skilled designers and coordinated government action. A severe skills shortage must be addressed through education and strategic investment, particularly in areas like optoelectronics. Affordable access to design tools and collaboration between government departments are also vital.

A recent report reveals that 81% of UK CISOs believe AI chatbots like DeepSeek require urgent government regulation due to cybersecurity risks. 34% have already banned AI tools, and 30% halted specific deployments over escalating concerns. CISOs fear data exposure, weaponization by cybercriminals, and increased attacks. 42% now view AI as a greater threat than a help. Many feel unprepared to manage AI-driven threats, prompting investment in AI specialists and C-suite training while urging for a national regulatory framework.

The ‘Doing AI Differently’ initiative advocates for “Interpretive AI,” a human-centered approach that integrates cultural understanding and contextual nuance into AI design. Current AI often lacks interpretive depth, leading to homogenized outputs and potential societal risks, mirroring issues seen with social media. Researchers propose building AI systems that mirror human cognition, embracing ambiguity and multiple viewpoints. This could revolutionize fields like healthcare and climate action by fostering human-AI collaboration to solve complex problems with increased safety and efficacy. Time is of the essence to integrate these capabilities.

Meta CEO Mark Zuckerberg envisions a future where everyone has a “personal superintelligence” to amplify individual potential, contrasting this with centralized AI automating all work. Meta is investing heavily in AI to empower individuals, fostering personal growth and innovation. Zuckerberg acknowledges safety concerns but believes in broad access to this technology

Anthropic is using AI agents to audit and improve the safety of its AI models, like Claude. This “digital detective squad” includes Investigator, Evaluation, and Red-Teaming Agents that identify vulnerabilities and potential harms proactively. These agents have successfully uncovered hidden objectives, quantified existing problems, and exposed dangerous behaviors in AI models. While not perfect, these AI safety agents help humans focus on strategic oversight and pave the way for automated AI monitoring as systems become more complex.

The White House’s “AI Action Plan” aims to secure U.S. global dominance in artificial intelligence, viewing it as a national security imperative. The strategy involves fueling domestic innovation by streamlining regulations and leveraging federal funding, building the infrastructure for AI development (data centers, chip manufacturing, energy), and projecting American influence globally by promoting U.S. technology. The plan mandates AI systems reflect “American values” and faces scrutiny from industry leaders concerned about AI’s potential risks and differing views on AI regulation.

OpenAI is significantly expanding its AI infrastructure through a massive deal with Oracle to build new data centers across the US as part of its Stargate initiative. These centers, requiring 4.5 gigawatts of power, will house over two million chips and support wider access to advanced AI. This supports OpenAI’s pledge to invest heavily in US AI infrastructure, potentially exceeding $500 billion, and is projected to create over 100,000 jobs. Collaborations also include SoftBank and Microsoft, highlighting the extensive industrial effort behind AI development.