Anthropic Claude Sonnet leads AI benchmark as Google lags

Key Takeaways

• Block CEO Jack Dorsey attributed layoffs of nearly half the company to the need to adapt to AI's future impact on corporate structures.
• The White House broke a contract with Anthropic, designating it a "supply chain risk" after the company refused to allow its AI system, Claude, for domestic mass surveillance.
• Microsoft reports that cybercriminals are increasingly using AI for tasks like creating phishing emails, debugging malware, and generating code, with North Korean groups employing AI for fake IT job applications.
• Better.com launched Tinman, an AI application developed with OpenAI, capable of approving home loans in seconds by analyzing a decade of financing data.
• A new benchmark, BullshitBench v2, revealed Anthropic's Claude Sonnet 4.6 identified nonsensical prompts over 90% of the time, outperforming OpenAI and Google models significantly.
• Jim Cramer suggested the weak February jobs report, particularly in the information sector, indicates the beginning of AI-related job losses, as companies adopt AI for tasks previously done by junior employees.
• The White House is urging states like Utah and Florida to delay their AI legislation, advocating for a unified federal framework instead of state-led efforts.
• Palo Alto Networks is expanding its AI-driven security solutions for operational technology and private 5G networks through partnerships with companies like Siemens, Nokia, and Aeris.
• Lenovo showcased friendly AI robots and AI-powered devices at MWC Barcelona 2026, highlighting an industry trend towards more natural, conversational AI interactions.
• CIQ introduced the CIQ Linux Kernel (CLK), an enterprise-grade kernel built on Linux 6.12 LT, designed to maximize the performance of modern AI hardware.

AI interviews person about life during wartime

An individual used AI to conduct an interview about their daily life and emotional state during a war. They describe feeling safe most of the time, with brief moments of danger when air sirens sound. The person also shares how their life has adapted to wartime, becoming more home-centered, and reflects on the emotional duality of fear and a sense of validation. The interview highlights that life continues with ordinary routines, punctuated by war reminders, and touches on spiritual reflections during this challenging period.

Jack Dorsey explains massive Block layoffs driven by AI

Block CEO Jack Dorsey explained his decision to lay off nearly half the company, stating it was a proactive move to adapt to the future of AI. He believes AI tools will dramatically change how companies are structured. Dorsey defended the layoffs as compassionate, offering generous severance packages. He emphasized that the shift is profound and that companies must rethink their structure to become more like an intelligence, moving away from traditional hierarchies.

Former Trump advisor criticizes White House AI contract decision

A former AI policy advisor to the Trump White House criticizes the Biden administration's decision to break a contract with AI company Anthropic. The Department of Defense considered Anthropic a supply chain risk, a designation never used against an American company before. This action followed Anthropic's refusal to allow the use of its AI system Claude for domestic mass surveillance of Americans. The advisor argues this move against a leading AI company is dangerous and sets a precedent for future AI use in warfare.

Michael Pollan explores consciousness and AI

Author Michael Pollan discusses the complex nature of consciousness in a conversation about his new book. He explains that certainty about consciousness is difficult, even for oneself, and relies on inference. Pollan is drawn to theories that focus on feelings as the origin of consciousness, rather than just thought processes. He notes the wide variety of theories attempting to explain consciousness and acknowledges the ongoing scientific and philosophical debate surrounding the topic.

North Korea uses AI for fake job applications, Microsoft reports

Microsoft reports that North Korean agents are using AI to enhance their scams by creating fake identities for remote IT job applications. These agents use AI tools like voice changers and face-swapping software to appear as credible Western candidates. Once hired, they funnel wages back to North Korea and have even threatened to leak company data. Microsoft has identified these groups as Jasper Sleet and Coral Sleet, noting their sophisticated use of AI to bypass security and maintain employment.

Microsoft: Threat actors use AI to boost cyberattacks

Microsoft reports that cybercriminals are increasingly using AI to improve their attacks, a practice they call 'operationalizing AI.' Threat actors use AI language models to create phishing emails, debug malware, and generate code, speeding up their operations. North Korean groups like Jasper Sleet and Coral Sleet are using AI for fake IT job applications and to maintain employment. Microsoft is also observing early experiments with agentic AI, where models can make decisions and execute tasks, potentially leading to more adaptive cyber threats.

Palo Alto Networks partners for AI security in 5G and OT

Palo Alto Networks is expanding its focus on AI-driven security for operational technology (OT) and private 5G networks. They announced a new industrial cybersecurity solution with Siemens for manufacturing and revealed collaborations with companies like Nokia and Aeris. These partnerships aim to provide AI-powered protection for industrial private 5G, data centers, and edge environments. The initiative seeks to secure sovereign AI and distributed edge use cases.

White House clashes with red states over AI laws

The White House is scrutinizing AI legislation proposed by Republican lawmakers in several red states, urging them to wait for a federal framework. States like Utah and Florida have seen their AI bills stalled or face pushback due to White House pressure. Lawmakers argue that state-led efforts are necessary to protect citizens regarding issues like children's safety and privacy. This tension highlights a growing conflict between state-level AI regulation and the federal government's desire for a unified approach.

AI and automation are changing R&D lab design

The design of modern research and development labs is being reshaped by the evolution of artificial intelligence and automation. Traditional lab spaces, built on older assumptions, are being re-evaluated to support new research workflows. As computational scientists, engineers, and automated systems become more integral to research, organizations are rethinking lab structures to foster both focused work and interdisciplinary collaboration.

Jim Cramer sees AI causing job losses in weak payroll report

CNBC's Jim Cramer believes a weak February jobs report, showing job losses in the information sector, signals the beginning of AI-related job cuts. He noted a significant decline in information sector jobs, attributing it to companies adopting AI for tasks previously done by junior employees. Cramer also pointed to job losses in transportation and warehousing, suggesting increased use of robots. He argues these are secular, not cyclical, changes impacting the economy.

Lenovo showcases friendly AI robots at MWC

Lenovo is leading a trend at MWC Barcelona 2026 to make AI more approachable with friendly robots and AI-powered devices. They presented concepts like the AI Workmate and an AI-powered laptop designed for more natural, conversational interactions. The event highlighted a broader industry shift towards humanizing AI, with many companies showcasing robots and devices that prioritize voice commands and natural language prompts over traditional screen interfaces.

Better.com uses AI for instant home loan approvals

Better.com has launched an AI application called Tinman, developed with OpenAI, that can approve home loans in seconds. This technology uses a decade of financing data and past loan decisions to analyze and underwrite mortgages, aiming to eliminate the weeks-long process. Better claims this will significantly lower interest rates for homebuyers by reducing origination costs. The app allows loan officers to process loans instantly, potentially challenging competitors like Rocket Mortgage and United Wholesale Mortgage.

New benchmark tests AI's ability to spot nonsense

A new benchmark called BullshitBench v2 tests AI models on their ability to identify nonsensical prompts that sound plausible. The benchmark uses domain-crossing jargon to create confusing questions. Anthropic's AI models, particularly Claude Sonnet 4.6, performed best, correctly identifying nonsense over 90% of the time. OpenAI and Google models lagged significantly, with newer versions showing little improvement. The study suggests that the drive for AI to be helpful might hinder its ability to recognize when a question is illogical.

CIQ launches Linux Kernel for AI hardware performance

CIQ has introduced the CIQ Linux Kernel (CLK), an enterprise-grade kernel designed to maximize the performance of modern AI hardware. CLK continuously tracks upstream Long Term kernel branches, providing the latest hardware enablement, performance optimizations, and security fixes. This ensures that enterprises can fully utilize their AI infrastructure without sacrificing stability or compatibility. CLK is built on Linux 6.12 LT and aims to close the gap between advanced hardware and production environments.

Sources

• The Blogs: An AI Interview: Our War, Up Close and Hopeful
• Jack Dorsey Is Ready to Explain the Block Layoffs
• Video: Opinion | Who Should Control A.I.?
• Is AI conscious? Michael Pollan weighs in on the debate
• North Korean agents using AI to trick western firms into hiring them, Microsoft says
• AI as tradecraft: How threat actors operationalize AI
• Palo Alto Networks Targets OT And 5G With New AI Security Alliances
• White House puts red state AI laws under scrutiny
• As AI evolves, the modern R&D lab is changing
• 'It's the beginning' — Cramer warns the weak payroll report shows AI-related job losses are here
• Lenovo Leads Push at MWC to Humanize AI With Friendly Robots
• Better says AI app OKs home loans in seconds, challenging Rocket, UWM
• BullshitBench Tests AI Models On Their Ability To Detect Plausible-Sounding Nonsense Prompts
• CIQ Introduces the CIQ Linux Kernel, Built to Unlock the Full Performance of Modern AI Hardware in Enterprise Production