Living Security is pushing organizations to adopt preemptive controls for AI agents before they cause incidents. Its Human Risk Management approach gives security teams visibility into both human behavior and agent activity, helping them spot risk early. The method connects to existing security tools and delivers targeted interventions based on who builds, prompts, or supervises agents.
AI security is drawing significant investment and attention. Sentra raised $150 million to address data security challenges created by AI agents. Meanwhile, CrowdStrike introduced an AI Kill Switch and announced plans to acquire Fiddler AI and Dreadnode, reflecting growing vendor interest in agent-level protections.
In the hardware space, about 600,000 Nvidia H200 chips are sitting idle in Hong Kong as Beijing delays approval for large-scale imports to mainland China. Despite rising demand, Chinese authorities have only allowed smaller shipments of roughly 10,000 chips each to major internet companies. An eight-card H200 server sold for over 6 million yuan by October, about three times its price a year earlier.
Tencent shares fell 2.46% after a rally driven by its new Hunyuan Hy4 AI model. The model adds visual understanding and computer-use features with 770 billion parameters, up from fewer than 300 billion in the Hy3 series. Hy4 is integrated into Tencent's WorkBuddy and CodeBuddy tools and sold through Tencent Cloud.
State pension funds hold at least $578.1 billion in AI-related companies, with the upper estimate reaching $706.6 billion as of June 30, 2026. NVIDIA was the largest single holding at $47.3 billion, followed by Apple and Microsoft. Researchers at the Equable Institute studied 51 large AI-related companies for the report.
Databricks executive Arsalan Tavakoli-Shiraji pushed back on warnings about overbuilt AI infrastructure, saying demand is almost infinite rather than reaching a limit. His comments contrast with market volatility that erased about $2.5 trillion in value. Federal Reserve researchers note that heavy spending on AI, data centers, and chips has boosted business investment even as data centers remain unpopular with the public.
Greek-founded AI startups are attracting international investors. Archiboost AI, based in San Francisco, raised $6.2 million from Greek and foreign investors including Marathon Venture Capital. The company uses AI to review architectural and engineering plans, detecting mistakes and coordination problems in construction projects.
AT&T updated its Connected Life security platform with Google Gemini AI features, including natural-language video search and daily Home Brief summaries. The system now offers 24/7 continuous recording and proximity automation. Professional monitoring costs $29.99 per month, while starter kits begin at $399.
Key Takeaways
- Living Security launched a Human Risk Management approach to help security teams control AI agents before incidents occur.
- Sentra raised $150 million to address data security challenges driven by AI agents.
- About 600,000 Nvidia H200 chips are stalled in Hong Kong awaiting Beijing's approval for shipment to mainland China.
- Tencent's new Hunyuan Hy4 AI model has 770 billion parameters and is integrated into WorkBuddy and CodeBuddy tools.
- US state and local pension funds hold at least $578.1 billion in AI-related companies, with NVIDIA as the largest holding at $47.3 billion.
- Databricks says AI demand is almost infinite, countering warnings about overbuilt infrastructure.
- Archiboost AI raised $6.2 million to use AI for reviewing architectural and engineering construction plans.
- AT&T added Google Gemini AI to its Connected Life home security cameras with natural-language video search.
- CrowdStrike introduced an AI Kill Switch and announced plans to acquire Fiddler AI and Dreadnode.
- AI investors may rely on infrastructure data that is up to two years old when valuing assets.
Living Security Urges Preemptive Controls as AI Agents Enter the Workforce
Living Security says security leaders must add controls for AI agents before they cause incidents. The company launched a Human Risk Management approach that gives teams data on both people and agents. It helps security teams spot risk early and act before problems happen. The approach connects to existing security tools already used by organizations. CEO Ashley Rose says strong controls plus human and agent visibility help security move from reactive to preemptive.
Living Security Says Human and Agent Risk Visibility Must Precede AI Controls
Living Security reports that security leaders are adding controls for AI agents. The company says Human Risk Management gives teams the data to act before an agent or a person creates an incident. The approach identifies risk across people, teams and AI agents. It also delivers targeted interventions and shows leadership how risk changes over time. Living Security works alongside the security tools teams already use.
Living Security Pushes Human Risk Management as AI Agents Join Workforces
Living Security says AI agents now draft, code, query data and take action inside enterprises. Security vendors are responding by adding agent inventory, access controls, monitoring and containment. CrowdStrike recently introduced an AI Kill Switch and announced plans to acquire Fiddler AI and Dreadnode. Living Security's Human Risk Management approach focuses on visibility into the people who build, prompt and supervise agents. CEO Ashley Rose says strong controls plus human and agent risk visibility help security teams move from reactive to preemptive.
Tencent Shares Fall 2.46% After Hunyuan AI Rally as Hong Kong Tech Cools
Tencent shares slipped 2.46% after a rally driven by its new Hunyuan AI model. The Hy4 model adds visual understanding and computer-use features with 770 billion parameters, up from fewer than 300 billion in the Hy3 series. Hy4 is already built into Tencent's WorkBuddy and CodeBuddy tools and sold through Tencent Cloud. The dip reflected the U.S. tech selloff, profit-taking and reports that about 600,000 NVIDIA H200 chips are stuck in Hong Kong awaiting Beijing's approval. Tencent trades at about 15 times earnings, which looks cheap next to U.S. AI leaders.
About 600,000 Nvidia H200 Chips Sit Idle in Hong Kong as China Delays Imports
About 600,000 Nvidia H200 chips are stored in Hong Kong awaiting shipment to mainland China. Beijing is withholding approval for large-scale imports despite rising demand from tech companies. China has allowed some chips to enter, with several major internet companies each obtaining about 10,000 H200 chips in recent months. But large-volume imports have not been approved. An eight-card H200 server was offered for more than 6 million yuan in mainland China by the end of October, about three times its price a year earlier. The H200 is not Nvidia's newest chip but has capabilities that Chinese alternatives have yet to match.
SAVeR squared improves safety for reasoning AI models
Researchers created SAVeR squared, a new safety method for Large Reasoning Models. It uses four fine-grained verifiable reward signals to check safety. The method removes the safety tax that usually hurts reasoning performance. Tests on models from 1.5B to 671B parameters showed better safety. It also kept general reasoning skills strong.
Sentra raises 150 million dollars for AI data security
Sentra raised 150 million dollars as AI agents create new data security challenges. The company focuses on protecting sensitive information from AI-driven threats. This funding reflects growing investor concern about data safety in the age of AI agents. The investment will support further development of security tools. These tools aim to keep data safe as AI usage grows.
AI leads investors to rely on outdated infrastructure data
AI investors are using information that is often two years old to value infrastructure. Brookfield, Macquarie, and KKR raised nearly 300 billion dollars in 2025. Physical AI startups raised 47.4 billion dollars in the first half of 2026. A market drop erased about 2.5 trillion dollars in value. The article notes that US bridges are inspected every two years, which can miss hidden risks. Dynamic Infrastructure is one company using AI to better check asset conditions across 15 US states.
Greek AI startups attract international investors
Greek-founded AI startups are gaining attention from investment funds. Archiboost AI, based in San Francisco, raised 6.2 million dollars from Greek and foreign investors including Marathon Venture Capital. The company built an AI platform that reviews architectural and engineering plans. It detects mistakes, omissions, and coordination problems in construction projects. This shows growing global interest in Greek AI innovation.
State pension funds hold billions in risky AI stocks
US state and local pension funds hold at least 578.1 billion dollars in AI-related companies. The upper estimate reached 706.6 billion dollars as of June 30, 2026. NVIDIA was the largest single holding at 47.3 billion dollars, followed by Apple and Microsoft. The Equable Institute studied 51 large AI-related companies for the report. When AI stocks fell, taxpayers could end up covering pension shortfalls since most public pensions are defined-benefit plans.
Data centers drive economic growth despite unpopularity
Federal Reserve researchers note that heavy spending on AI, data centers, chips, and networking gear has boosted business investment and financial markets. Experts say data centers remain unpopular with the public even though they play a key role in economic growth. The trend reflects rising demand for AI technology across many industries.
Databricks says AI demand is almost infinite
Databricks executive Arsalan Tavakoli-Shiraji says innovation is accelerating despite market warnings about overbuilt AI and possible write-offs. He believes demand for AI is almost infinite rather than reaching a limit. His comments suggest confidence that AI capacity will continue to grow.
AT&T adds Google Gemini AI to home security cameras
AT&T updated its Connected Life security platform with Google Gemini AI features, including natural-language video search and daily Home Brief summaries. The system now offers 24/7 continuous recording and proximity automation that arms or disarms alarms as people walk through the door. Professional monitoring costs $29.99 per month, while hardware starter kits start at $399 and go up to $699 for multi-camera setups.
Sources
- AI Agents Are Joining the Workforce - Living Security Says Preemptive Security Starts With Human and Agent Risk Visibility
- AI Agents Are Joining the Workforce - Living Security Says Preemptive Security Starts With Human and Agent Risk Visibility
- AI Agents Are Joining the Workforce - Living Security Says Preemptive Security Starts With Human and Agent Risk Visibility
- Tencent Stocks Slip 2.46% After Hunyuan AI Rally as Hong Kong Tech Cools
- Nvidia chips pile up in Hong Kong as China withholds import approval
- SAVeR$^2$: Reasoning-based Safety Alignment for Large Reasoning Models via Verifiable Rewards
- Sentra raises $150m as AI agents challenge data security
- AI: Investors Are Pricing Infrastructure on Two-Year-Old Information
- Greek-made AI instruments gain investors’ attention
- State Pension Funds Have at Least $578 Billion Riding on Tumbling AI Stocks
- Data centers are unpopular — and a key motor of economic growth, experts say
- AI demand is 'almost infinite', not worried about too much capacity: Databricks
- AT&T Brings Google Gemini Intelligence to Connected Life Security Cameras
Comments
Please log in to post a comment.