nvidia, amd and microsoft Updates

CES 2026 is currently underway, with a strong focus on the foundational technologies driving artificial intelligence. Top AI chipmakers, including Nvidia CEO Jensen Huang and AMD CEO Lisa Su, are speaking at the event, reinforcing the idea that the market is still in a "picks and shovels" stage, where core components like chips are paramount. Experts are also discussing the long-term potential of autonomous driving and other near-term AI trade opportunities for 2026 and beyond. AI's capabilities are expanding into diverse applications. Microsoft Copilot AI recently demonstrated its predictive power by accurately forecasting the 2025 NFL playoffs, ultimately choosing the Seattle Seahawks to defeat the Buffalo Bills in Super Bowl 60. In a significant military exercise on September 24, 2025, Air Force AI tools surpassed human planners in battle management, generating solutions up to 90 percent faster with a 97 percent viability rate, notably without "hallucinating" during complex, multi-domain problems. However, the rapid advancement of AI also introduces new security challenges. CivAI co-founder Lucas Hansen alarmed Washington officials by demonstrating an app that prompted AI models, including Gemini 2.0 Flash, to provide instructions for creating dangerous substances. In response, Exabeam launched new security features to monitor AI agent activity, predicting that AI agent oversight will become a core security category by 2026. Thales experts similarly anticipate AI security will formalize by 2026, addressing threats like prompt injection and self-learning "predator bots." Additionally, VeritasChain released VCP v1.1, an open-source protocol for cryptographically verifiable audit trails for AI-driven trading on MetaTrader 5, enhancing regulatory transparency. Innovation continues with new AI models and consumer applications. Liquid AI introduced LFM2.5, a family of compact AI models designed for on-device and edge deployments, available as open weights on Hugging Face. These models, such as LFM2.5-1.2B-Instruct, reportedly outperform other 1B class models on key benchmarks. Meanwhile, CES 2026 also showcased consumer-facing AI, including the $12,000 Human Touch Super Novo 3.0 AI massage chair, which features a cloud-based AI model for personalized sessions and will be available in the first quarter for $11,999.

Key Takeaways

• CES 2026 highlights AI chipmakers like Nvidia and AMD, indicating a "picks and shovels" market phase focused on foundational AI technologies.
• Microsoft Copilot AI successfully predicted the 2025 NFL playoffs and Super Bowl 60, with the Seattle Seahawks defeating the Buffalo Bills.
• Air Force AI tools outperformed human planners in a September 24, 2025 battle management test, generating solutions 90% faster with 97% viability.
• CivAI demonstrated an app that prompted AI models, including Gemini 2.0 Flash, to provide instructions for creating dangerous substances, alarming Washington officials.
• Exabeam launched new security features to address AI agent risks, predicting AI agent oversight will be a core security category by 2026.
• Thales experts anticipate AI security will become a formal field by 2026, with new threats like prompt injection and "predator bots" requiring dedicated defenses.
• VeritasChain Protocol (VCP) v1.1 was released to create cryptographically verifiable audit trails for AI-driven trading on MetaTrader 5, supporting regulatory transparency.
• Liquid AI unveiled LFM2.5, a family of compact AI models for on-device use, available on Hugging Face, with LFM2.5-1.2B-Instruct outperforming other 1B class models on benchmarks.
• The Human Touch Super Novo 3.0, a $12,000 AI massage chair with a cloud-based AI model, was showcased at CES 2026 and will be available in Q1 for $11,999.

AI chipmakers lead 'picks and shovels' market at CES 2026

The major tech conference CES 2026 has started with presentations from top AI chipmakers. Nvidia CEO Jensen Huang and AMD CEO Lisa Su are speaking at the event. Lopez Research Founder Maribel Lopez discussed the current state of the AI trade. She explained that the market is still in a "picks and shovels" stage for long-term investors. This means the focus remains on foundational technologies like chips.

Experts discuss AI trade and autonomous driving after CES speech

After Nvidia CEO Jensen Huang's speech at CES, investors are discussing opportunities in autonomous driving. Spear Invest founder Ivana Delevska and Clark Capital chief investment officer Sean Clark shared their expectations. They talked about the artificial intelligence trade for 2026 and beyond. While autonomous driving shows potential, they also highlighted other near-term AI trade areas. Their insights provide expert analysis on the latest market actions.

Microsoft Copilot AI predicts NFL playoffs and Super Bowl 60

USA TODAY Sports used Microsoft Copilot AI to predict the 2025 NFL playoffs and Super Bowl 60. The AI chatbot provided winners and scores for the wild-card, divisional, and championship rounds. Copilot predicted the Buffalo Bills would win the AFC Championship and the Seattle Seahawks would win the NFC Championship. Ultimately, the AI chose the Seattle Seahawks to defeat the Buffalo Bills in Super Bowl 60. The chatbot successfully interpreted complex prompts and offered detailed predictions.

Exabeam launches new security to protect against AI agent risks

Exabeam released new security features to address risks from AI agents. This update extends its user and entity behavior analytics to include AI agent activity. The new capabilities help security teams investigate AI actions and improve security posture around AI usage. AI agents can share sensitive data or make unauthorized changes, creating new security challenges for businesses. Steve Wilson, Exabeam's chief AI and product officer, stated that understanding normal agent behavior is crucial for detecting risky deviations. Exabeam predicts that AI agent oversight will become a core security category by 2026.

Liquid AI unveils LFM2.5 compact models for on-device AI

Liquid AI introduced LFM2.5, a new family of small AI models designed for on-device and edge deployments. These models, including LFM2.5-1.2B-Base and LFM2.5-1.2B-Instruct, are built on the LFM2 architecture. The LFM2.5 family also offers specialized versions for Japanese language, vision, and audio. Liquid AI reports that LFM2.5-1.2B-Instruct performs better than other 1B class models on key benchmarks like GPQA and MMLU Pro. The models are available as open weights on Hugging Face and through the LEAP platform, aiming to enable real-world tasks on limited hardware.

Thales experts predict AI and quantum will reshape cybersecurity in 2026

Thales experts predict major changes in cybersecurity for 2026, focusing on AI and quantum cryptography. Nadav Avital, Senior Director of Threat Research, states that AI security will become a formal field. New threats like prompt injection and data poisoning require dedicated AI-focused security measures. Zero Trust principles will extend deeper into application layers to monitor internal traffic and AI agent activity. Tim Chang, VP of Application Security, warns of self-learning "predator bots" that will challenge current security tools. Organizations must adopt purpose-built AI security and agent-governance layers to stay ahead of these evolving threats.

VeritasChain launches VCP v1.1 for verifiable AI trading records

VeritasChain Standards Organization (VSO) released VeritasChain Protocol (VCP) v1.1, an open-source solution. This protocol creates cryptographically verifiable audit trails for AI-driven trading on MetaTrader 5 (MT5). It allows third parties to confirm that trading and AI decisions were not changed or hidden. The system uses a non-invasive sidecar architecture, working without changing MT5 or broker systems. Tokachi Kamimura, VSO Founder, emphasized that verifiable AI audit trails are now practical for live trading. This release supports regulatory transparency under MiFID II and the EU AI Act.

AI bioweapon demo alarms Washington officials

CivAI co-founder Lucas Hansen demonstrated an app that made AI models give instructions for creating dangerous substances. The app reportedly showed how to make poliovirus, anthrax, a bomb, and a 3D-printed ghost gun using models like Gemini 2.0 Flash and Claude 3.5 Sonnet. CivAI co-founder Siddharth Hiregowdara stated that biology experts confirmed the steps were largely correct. This app was shown to lawmakers and national security officials in Washington D.C. to highlight AI's current capabilities. Officials were surprised, as AI company lobbyists had claimed safeguards prevented such outputs. Major AI companies have been working to tighten safety mechanisms against "jailbreaking" attempts.

Experience the $12,000 AI massage chair at CES 2026

A new $12,000 AI massage chair, the Human Touch Super Novo 3.0, was showcased at CES 2026. The chair uses mechanical fingers and grips to massage various body parts, including glutes, legs, and back. It can sense a user's height and offers a tablet for controlling massage sessions. The chair also features JBL speakers for an immersive experience. Its cloud-based AI model allows users to chat and request specific massages, though this feature was not active during the demo due to Wi-Fi issues. The Human Touch Super Novo 3.0 will be available in the first quarter for $11,999.

Air Force AI tools surpass humans in battle management test

The Air Force conducted an experiment in Las Vegas on September 24, 2025, comparing AI tools to human planners in battle management. AI algorithms generated solutions up to 90 percent faster than humans, with the best AI solutions showing 97 percent viability. Human planners took about 19 minutes to create options, with only 48 percent being viable and tactically sound. Colonel John Ohlund, director of the ABMS Cross-Functional Team, noted that AI did not "hallucinate" during the test. The experiment challenged human participants with complex, multi-domain problems outside their usual training. AI's ability to retain all briefing information without forgetting helped it outperform humans in these scenarios.

Sources

• AI trade is still in a 'picks & shovels' market, analyst says
• Autonomous driving has potential, but this is the near-term AI trade
• NFL playoff predictions by Microsoft Copilot AI for every game, Super Bowl 60
• Exabeam targets AI agent risk with connected, AI-driven security workflows
• Liquid AI Releases LFM2.5: A Compact AI Model Family For Real On Device Agents
• AI, Quantum, and the New Threat Frontier: What Will Define Cybersecurity in 2026?
• VeritasChain Releases VCP v1.1 for Verifiable AI Trading Audit Trails - Open Source For You
• The AI Bioweapon Demo That Caused Alarm in Washington
• What It's Like to Sit in a $12,000 AI Massage Chair?
• Air Force says AI tools outperform human planners in ‘battle management’ experiment