Nvidia investigates Huang over alleged spy contact, DeepSeek's agent tool hits 1.5M users
Nvidia's board has opened an investigation into CEO Jensen Huang's ties to an executive reportedly recruited by Chinese intelligence, while DeepSeek's coding agent framework posts rapid growth and Claude turns up inside malware.
Every line links to a primary source in the full briefing.
01Nvidia opens investigation into Jensen Huang over alleged Chinese intelligence contact
Nvidia's board has launched an independent investigation into CEO Jensen Huang following reports that he had a close, reportedly romantic, relationship with a Megaspeed executive who US officials believe was recruited by China's Ministry of State Security. The board says it has no current evidence that national security or company secrets were compromised, but an outside counsel-led committee will review the matter.
A leadership investigation at the company that supplies most of the chips behind AI infrastructure is worth watching even from a distance, since it introduces uncertainty around a supplier many AI tools and cloud services depend on.
Source: The Verge
02DeepSeek's agent framework surges to 1.5 million daily users
DeepSeek's Harness coding and agent framework reached 1.54 million daily active users within three months of release, with reported daily token throughput exceeding 92 trillion. Separately, DeepSeek said it had fixed a disruption affecting its chat service and two API endpoints.
Rapid uptake of a free, open alternative gives smaller firms another low-cost option for building AI-powered tools, reducing reliance on a handful of paid providers.
03Malware uses Claude to automatically profile and rank stolen accounts
Cisco researchers reported finding malware that calls on Anthropic's Claude to automatically analyse stolen account data and rank which victims are most valuable to target further. The technique shows attackers embedding mainstream AI models directly into automated criminal workflows rather than using custom-built tools.
Any business storing customer or account data should assume attackers now have AI help triaging what they steal, which raises the cost of a breach and the case for tighter access controls.
Source: The SOFX Report
04Spanish firms expect to hire more tech staff as AI adoption grows
A survey covered by RRHH Digital finds eight in ten companies in Spain expect to hire more technology profiles because of rising AI adoption. Separately, a regional report on Extremadura found only about one in four businesses there currently use AI, roughly ten points below other areas, pointing to uneven adoption across the country.
Competition for AI-literate staff is likely to intensify even in regions where adoption is still low, so waiting too long to build internal skills could get more expensive.
05Web Summit opens in Lisbon with AI and robotics centre stage
Web Summit 2026 has opened in Lisbon, with AI and robotics framed as central themes alongside broader discussion of Europe's digital future. The event draws companies and policymakers debating how the region positions itself in the global AI race.
Europe-focused conferences like this are often where new partnerships, funding announcements and policy signals first surface, so they're worth a glance even for owners not attending.
Source: Essential Business
06Debate over an AI valuation bubble resurfaces as Anthropic's IPO looms
Commentary pieces this week argue over whether sky-high AI company valuations, including Anthropic's reported IPO target, reflect a genuine bubble or justified growth. One widely shared essay pushes back on the assumption that a collapse is inevitable, while other coverage ties Anthropic's IPO directly to broader risk in the AI boom.
Owners relying on AI vendors for core tools should factor in that a sharp correction in AI company valuations could disrupt pricing, support or continuity of services they depend on.
Source: Yahoo FinanceHacker News
Review which AI tools touch customer data and who can access the logs
With malware now using mainstream AI models to profile and rank stolen accounts, it's a good moment to check who inside your business, and which connected tools, can see customer account data. Tightening access and reviewing logs this week costs little and reduces what an attacker could exploit if a breach happens.
Sources
9 links- 01DeepSeek's Harness Framework reaches 1.54 million DAU in three months, with daily token usage exceeding 92 trillion.KuCoin
- 02Ocho de cada diez empresas en España contratarán más perfiles tecnológicos por el auge de la IARRHH Digital
- 03Extremadura: 1 de cada 4 empresas usa IA, 10 puntos menosEl Ecosistema Startup
- 04Web Summit 2026 opens in Lisbon with AI, robotics and Europe’s digital future in the spotlightEssential Business
- 05DeepSeek Applies Fix for Disruption to Chat and Two API ServicesTOKENPOST
- 06Anthropic’s $1.5 Trillion IPO Could Expose the Biggest Risk Facing the AI BoomYahoo Finance
- 07Cisco Finds Malware Using Claude AI to Profile Victims and Rank Stolen AccountsThe SOFX Report
- 08Nvidia is investigating Jensen Huang’s relationship with a Chinese spyThe Verge
- 09I'm not convinced the AI bubble existsHacker News