Intel and AMD detail AI chip plans as OpenAI adds Codex Security

Key Takeaways

• Intel and Altera are developing sensor-bridge designs using Intel systems and programmable Altera FPGAs for edge AI in healthcare, aerospace and defense, and robotics.
• Bloomberg reports that 10 AI-linked companies accounted for more than half of the S&P 500’s $32 trillion gain since ChatGPT’s release, complicating debates over AI regulation and economic growth.
• OpenAI’s Codex Security identifies vulnerabilities, proposes patches and checks fixes; its reported false-positive rate was 0.81%, with rollbacks below 1%.
• AI-safety researchers argue that governance should constrain agents’ direct control and access to knowledge, not intelligence alone.
• Alibaba Health’s survey found 93% of hundreds of Chinese doctors use AI tools, up from about 70% a year earlier; digital health marketing spending is forecast to reach 274.4 billion yuan ($41.1 billion) by 2030.
• AMD says Zen 6, codenamed Venice, was not designed specifically for agentic AI; development began before the recent surge in interest.
• Data-centre operators are adopting AI video analytics for threat detection and security monitoring as facilities expand to meet AI demand.
• Rafael unveiled an AI-assisted cruise missile carrying a 453-kilogram (1,000-pound) warhead, the heaviest payload publicly disclosed for an Israeli missile.
• Illumio joins WWT’s ARMOR AI risk framework, developed with NVIDIA and Texas A&M, with a focus on breach containment and segmentation.
• BCG’s five AI priorities for medtech sales are dynamic targeting, personalized engagement, AI-supported sales, data-driven pricing and margin optimization, and 24/7 service agents.

Intel and Altera Build Sensor Tools for Edge AI

Intel and Altera announced a collaboration on modular sensor bridge designs for edge and physical AI systems. The planned reference solutions combine Intel compute, connectivity and software with Altera FPGAs, which can be reprogrammed as sensors and AI models change. The designs aim to help equipment makers connect and process data from many sensors while meeting timing, power, performance and cost needs. The companies are developing solutions for healthcare, aerospace and defense, and robotics, and plan to offer them through Intel’s Open Edge Platform SDK and an Altera SDK.

Intel and Altera Plan Modular Sensor Bridges for Physical AI

Intel and Altera are developing modular reference solutions that connect sensors to computing systems used in edge and physical AI. The designs pair Intel’s compute, connectivity and software with Altera’s programmable FPGAs to help equipment makers handle different sensor interfaces and real-time processing needs. The work targets healthcare equipment, aerospace and defense mission systems, and robotics. The companies plan to make the solutions available through Intel’s Open Edge Platform SDK and an Altera SDK, but they have not described a finished commercial product.

Democrats Face a Challenge Balancing AI Rules and Economic Growth

The article says Democrats face a challenge as they consider regulating AI without harming the economy. After Democrats retook the House and Senate, the stock market fell sharply, and major technology companies lost value. AI has become closely tied to business spending and market growth, with a Bloomberg report saying more than half of the S&P 500’s $32 trillion gain since ChatGPT’s release came from 10 AI-linked companies. The article also notes that AI demand may contribute to inflation and higher long-term bond yields, while Democrats lack a clear plan for regulating the industry.

OpenAI Brings Codex Security Tools to the Cloud

OpenAI is making Codex Security available as a cloud product after testing security tools on its own work. The system uses custom scan prompts and a shared knowledge base to find vulnerabilities, then classifies flaws, creates patches and checks whether the fixes work. OpenAI reported that 36% of findings in its sprint were duplicates and that its false positive rate was 0.81%, with rollbacks below 1%. The Codex Security CLI is available as @openai/codex-security on NPM, and the Cloud product was expected in the plugin marketplace before the end of the week.

AI Safety Researchers Urge Limits on Control and Knowledge

The article proposes a framework called the competence circuit to study how AI agents could affect the real world through intelligence, know-how and direct control. Its simulations suggest that faster intelligence does not always mean rapid progress, because important tests and the time needed to gain knowledge can still limit an agent. The researchers also found that required authorization delays can slow highly capable agents. They argue that AI governance should focus on restricting control and setting knowledge boundaries, rather than limiting intelligence alone, and say physical-world checks may be important for digital systems.

Drugmakers turn to China’s AI apps to reach doctors

Drugmakers in China are exploring AI-powered medical apps as a new way to reach doctors after a graft crackdown made traditional outreach harder. Alibaba Health, JD Health and Yidu Tech are among the companies developing or discussing these tools, while medical education firm PICA Health has added JD Health’s Zhiyi assistant to its app. Alibaba Health is considering a model that would show promoted medical research alongside answers from its Hydrion app, though it has not started working with drugmakers. Alibaba Health found that 93% of hundreds of doctors it surveyed use AI tools, up from about 70% a year earlier. Frost & Sullivan estimates China’s digital health marketing spending will grow by an average of 27.5% per year from 2027 to 2030, reaching 274.4 billion yuan, or $41.1 billion.

AMD says Zen 6 was not designed specifically for agentic AI

AMD executives Mike Clark, the company’s chief Zen architect, and corporate vice president Madhu Rangarajan said Zen 6 was not designed specifically for agentic AI. The comments challenge descriptions from Nvidia and Arm that their latest data center CPUs are built for agentic AI. AMD said development of the Zen 6 family, codenamed Venice, began before the agentic AI wave took off around March this year. CPU development takes years, the executives noted, so chipmakers could not have planned for the recent surge in interest from the start.

AI video tools strengthen data centre security

Data centre operators are using AI-powered video and analytics to improve physical security as facilities expand to meet growing AI demand. These tools can work with access control and surveillance systems to help operators spot and respond to threats. Automation can also handle monitoring and response tasks, allowing security staff to focus on other work. The technology gives operators a broader view of their facilities and can help them identify risks more effectively. The article describes AI-powered video as a growing trend in data centre security.

Rafael unveils AI-assisted cruise missile with heavy warhead

Israeli state-owned defense contractor Rafael Advanced Defense Systems unveiled a new AI-assisted cruise missile on Tuesday. The missile can carry a 453-kilogram, or 1,000-pound, warhead. The article says this is the heaviest payload publicly disclosed for an Israeli missile to date.

Illumio joins WWT framework for managing AI security risks

Illumio will be included in World Wide Technology’s AI Readiness Model for Operational Resilience, known as ARMOR, which helps organisations assess and manage risks in AI environments. Illumio’s role focuses on breach containment, including mapping connections across workloads and devices and using segmentation to limit how far an attack can spread. The company says its controls cover data centres, cloud environments, containers, endpoints and operational technology, as well as connections to public AI services. ARMOR was developed by WWT with NVIDIA and the Texas A&M University System, and organizes AI security into seven domains, including infrastructure security, identity security and data security. Illumio said its controls can also extend to AI infrastructure using NVIDIA BlueField-3 DPUs.

BCG study outlines five AI moves for stronger medtech sales

Most medtech leaders in the United States and around the world rate their commercial skills as basic to intermediate. A few companies are ahead of the sector index by eight percentage points in annual shareholder returns. BCG’s study, based on more than 15 years of data and insights, says AI is the next step in improving commercial performance. It recommends five moves: dynamic customer targeting, personalized engagement, AI-supported and autonomous sales, data-driven pricing and margin optimization, and 24/7 service agents.

Sources

• Intel, Altera Partner on Sensor Integration for Edge and Physical AI
• Intel and Altera Develop Modular Sensor Bridges for Physical AI
• Democrats’ Next Big AI Challenge
• OpenAI turns defense factory into Cloud product
• AI Governance Should Prioritize Control and Knowledge Boundaries Over Limiting Intelligence
• Analysis-Drugmakers tap China AI app boom to reach doctors as crackdown upends marketing
• AMD VP and chief Zen architect rebuke claims that Nvidia and Arm made CPUs for agentic AI — ‘We didn’t magically know four years ago that agentic was going to be a thing’
• How AI-powered video is reshaping data centre physical security
• Israeli firm Rafael launches new AI-assisted cruise missile
• Illumio joins WWT's AI resilience framework for containment
• Modernizing the Milkman: AI’s New Playbook for Medtech Commercial Excellence