OpenAI tools reveal security flaws as Apple develops AI chips

The artificial intelligence sector is seeing significant developments across security, hardware, and application, with a clear trend towards on-device processing and a renewed focus on human authenticity. WitnessAI, an AI safety platform startup, launched its Agentic Security capability on January 13, 2026. This new tool helps security teams monitor, control, and protect autonomous AI agents accessing company systems, defending against threats like prompt injection and safeguarding sensitive data. The company also secured $58 million in new funding, led by Sound Ventures, to expand its global business and enhance its agentic AI security features. Simultaneously, concerns about AI security are rising. Research from Tenzai, an AI security startup, reveals that websites built using coding tools from OpenAI, Anthropic, Cursor, Replit, and Devin contain security vulnerabilities, including cross-site scripting and injection issues. Tenzai CEO Ben Perez stated these tools are not yet ready for widespread use, and his company is collaborating with the AI developers to address these weaknesses. CrowdStrike CEO George Kurtz also voiced concerns about the unpredictable nature of AI agents, even as his company actively acquires cybersecurity startups to strengthen its defenses. In hardware, Apple plans to begin mass producing its own AI server chips, codenamed Baltra, in the second half of 2026. This strategic move, developed with Broadcom, aims to bring more AI operations in-house, reducing reliance on external companies and ensuring tighter control over data processing and privacy. This comes even as Apple recently partnered with Google for Siri's AI. The broader industry is also exploring a shift towards smaller, local data centers, with Perplexity CEO Aravind Srinivas suggesting powerful AI tools will eventually run directly on device hardware. Microsoft's Copilot+ laptops and Apple Intelligence already utilize on-device AI for enhanced speed and security. RiseLink Technologies showcased the future of physical AI at CES 2026, highlighting the movement of AI from cloud systems to physical devices. CEO Dr. Pengfei Zhang emphasized the need for on-device intelligence and efficient design, exemplified by their new AI-powered early education toy, ChooChoo, which uses on-device AI for interactive reading and storytelling with a focus on privacy. Meanwhile, Evertune launched its Shopping Intelligence capability on January 14, 2026, helping brands track product visibility in AI-powered recommendations, particularly within ChatGPT shopping results that now feature direct purchase links. Beyond commercial applications, AI is proving beneficial for environmental protection and food security. The Nature Conservancy uses AI and electronic monitoring to improve fishing oversight and combat illegal fishing, while Microsoft's AI for Good Lab leverages AI to monitor agriculture, aiding food security decisions during crises. On the policy front, House members Gabe Amo and Gregory W. Meeks criticized President Trump's January 12 decision to ease export controls on advanced AI chips to China, citing national security risks. Amidst these technological advancements, a report from Billion Dollar Boy indicates a significant shift in consumer preference, with only 26% now favoring AI-generated content in 2026, down from 60% in 2023, underscoring a growing value for human authenticity and originality.

Key Takeaways

• WitnessAI launched Agentic Security on January 13, 2026, to secure autonomous AI agents and raised $58 million in funding led by Sound Ventures.
• Research by Tenzai indicates AI coding tools from OpenAI, Anthropic, Cursor, Replit, and Devin produce websites with security flaws like cross-site scripting.
• Apple plans to mass produce its in-house AI server chips, codenamed Baltra, by late 2026 to bring AI operations in-house and reduce external reliance.
• RiseLink Technologies showcased on-device AI at CES 2026, introducing ChooChoo, an AI-powered early education toy using local intelligence.
• Evertune launched Shopping Intelligence on January 14, 2026, to help brands track product visibility in AI-powered recommendations, including ChatGPT shopping results.
• AI and big data are being used by The Nature Conservancy and Microsoft's AI for Good Lab to enhance fishing oversight and monitor agriculture for food security.
• CrowdStrike CEO George Kurtz expressed concerns about unpredictable AI agents while actively acquiring cybersecurity startups.
• A Billion Dollar Boy report shows only 26% of consumers prefer AI-generated content in 2026, a significant drop from 60% in 2023, highlighting a shift towards human authenticity.
• Congress members criticized President Trump's January 12 decision to ease export controls on advanced AI chips to China, citing national security concerns.
• The industry sees a growing interest in smaller, local "edge" data centers, with Apple Intelligence and Microsoft's Copilot+ laptops already utilizing on-device AI.

WitnessAI launches Agentic Security for AI agents

WitnessAI, an AI safety platform startup, launched its new Agentic Security capability on January 13, 2026. This tool helps security teams see, control, and protect autonomous AI agents that access company systems and tools. Many companies use these agents, but security teams often do not know which agents exist or what they access. Agentic Security finds agent activity in popular AI environments and links agent actions back to the people who started them. It also defends against attacks like prompt injection and protects sensitive data.

WitnessAI raises 58 million dollars for AI security

WitnessAI, an AI security company based in Mountain View, California, secured $58 million in new funding. Sound Ventures led this investment round, with other participants including Fin Capital and Qualcomm Ventures. The company will use this money to grow its global business, expand its products, and improve its agentic AI security features. WitnessAI offers a platform that helps businesses secure all their AI interactions and autonomous agents. It provides clear visibility into AI activity and enforces smart policies to stop new threats like prompt injection.

RiseLink gathers AI leaders at CES 2026

RiseLink Technologies brought together leaders in AI, hardware, and connected products at CES 2026 in Las Vegas on January 14, 2026. The company, which makes edge intelligence and low-power connectivity solutions, discussed how AI is moving from cloud systems to physical devices. Dr. Diana Zhu and CEO Dr. Pengfei Zhang highlighted the need for on-device intelligence and efficient design for products like smart homes and robots. At the event, RiseLink also introduced ChooChoo, an AI-powered early education toy designed to help children with language development using on-device AI.

RiseLink shows physical AI future at CES 2026

RiseLink Technologies gathered leaders from AI, hardware, and connected products at CES 2026 in Las Vegas on January 15, 2026. The company, which provides edge intelligence and low-power connectivity, discussed how AI is moving from the cloud into physical devices. Dr. Pengfei Zhang, RiseLink's CEO, stated that the future of AI is physical and requires deep collaboration. They showcased ChooChoo, an AI-powered early education toy, as an example of this new approach. ChooChoo uses on-device intelligence for interactive reading and storytelling, focusing on privacy and safety.

Evertune launches Shopping Intelligence for AI recommendations

Evertune, a leader in Generative Engine Optimization, launched its new Shopping Intelligence capability on January 14, 2026. This tool helps brands track and improve how visible their products are in AI-powered recommendations, especially in ChatGPT shopping results. ChatGPT now shows product cards with direct purchase links, which changes how consumers find products. Shopping Intelligence captures these AI shopping experiences, showing brands when their products appear, how they compare to competitors, and which retailers are most effective. It helps brands understand what to promote, fill content gaps, and strengthen retail partnerships.

AI coding tools create websites with security flaws

New research from AI security startup Tenzai shows that websites built using coding tools from OpenAI, Anthropic, Cursor, Replit, and Devin have security problems. Tenzai CEO Ben Perez stated that these AI coding tools are not yet ready for widespread use, as developers might accidentally create vulnerable code. The flaws found include cross-site scripting and injection issues, which could allow attackers to steal private information or take over websites. Tenzai is now working with the AI companies to fix these security weaknesses.

AI helps protect nature and secure food supply

Big data and AI offer great potential to improve food systems and protect nature, both on land and in the ocean. Food systems worldwide face challenges like overfishing and climate change, which threaten livelihoods and global supply chains. The Nature Conservancy uses AI and electronic monitoring to strengthen fishing oversight, reduce illegal fishing, and improve returns for Pacific Island nations. Microsoft's AI for Good Lab also uses AI and other technologies to monitor agriculture, helping with food security decisions during natural disasters or conflicts. These efforts show how AI can help build more sustainable food systems for everyone.

Apple plans in-house AI server chips by late 2026

Apple plans to start mass producing its own AI server chips in the second half of 2026. This move aims to bring more AI operations in-house and reduce reliance on other companies. Analyst Ming-Chi Kuo reported that the new chip, codenamed Baltra, is being developed with Broadcom and is specifically designed for AI tasks. While Apple recently partnered with Google for Siri's AI, this internal chip development shows a dual strategy. Apple wants to gain tighter control over AI data processing and ensure its hardware meets its privacy and optimization standards.

Small data centers could change AI processing

The future of data centers might involve smaller, local units instead of huge remote facilities. Perplexity CEO Aravind Srinivas believes powerful AI tools will eventually run directly on device hardware, reducing the need for large data centers. Apple Intelligence and Microsoft's Copilot+ laptops already use on-device AI for speed and security. While many large data centers are still being built, there is a growing interest in smaller "edge" data centers that can provide heating as a byproduct. Mark Bjornsgaard, founder of DeepGreen, envisions every public building housing a small data center to create a vast, efficient network.

CrowdStrike CEO warns about unpredictable AI agents

George Kurtz, CEO of CrowdStrike, expressed concerns about the unpredictable behavior of AI agents. Despite these worries, his cybersecurity company is actively buying more cybersecurity startups to strengthen its defenses. CrowdStrike's strategy aims to keep the company ahead in the fast-changing world of cyber threats. Kurtz highlights that AI can be a strong tool for security but also brings new, unexpected risks when autonomous agents are involved. This aggressive acquisition plan helps CrowdStrike expand its platform and tackle complex, AI-driven attacks.

Human authenticity gains value over AI content

After 2025 saw a flood of AI-generated content, 2026 marks a shift where human authenticity is highly valued. A report from Billion Dollar Boy shows that only 26% of consumers now prefer AI-generated content, a big drop from 60% in 2023. This change means platforms must define authenticity, and creators need to highlight their unique human qualities and imperfections to stand out. Brands are also becoming more open to creators' "messiness" and may even ask for it in future deals. The message is clear: originality and genuine human creativity are now key differentiators.

Congress members criticize Trump AI chip sales to China

On January 12, House members Gabe Amo and Gregory W. Meeks criticized President Trump's decision to ease export controls on advanced AI chips sold to China. They voiced their concerns to Secretary of Commerce Howard Lutnick, arguing that this move could harm U.S. national security. The members believe that allowing these sales jeopardizes America's technological lead and could boost China's AI capabilities. They also suggested that collecting money from these sales might go against current export laws.

Sources

• WitnessAI debuts Agentic Security as enterprises deploy more autonomous AI agents
• WitnessAI Secures $58 Million in Strategic Funding
• RiseLink Convenes the Physical AI Ecosystem at CES 2026, Showcasing How Intelligence Moves From Silicon to Shipping Products
• RiseLink Convenes the Physical AI Ecosystem at CES 2026, Showcasing How Intelligence Moves From Silicon to Shipping Products
• Evertune Launches Shopping Intelligence to Track Brand Visibility in AI-Powered Product Recommendations
• Anthropic and OpenAI’s Coding Tools Produce Websites with Security Flaws
• From ocean to farm: harnessing AI for nature, food security
• Apple could unveil in-house AI server chips later this year to reduce reliance on partners
• Honey, I shrunk the data centres: Is small the new big?
• CrowdStrike CEO says AI agents are unpredictable as company snaps up more cybersecurity startups
• After an oversaturation of AI-generated content, creators’ authenticity and ‘messiness’ are in high demand
• Press Release: Amo and Meeks Critique Trump's Approval of AI Chip Sales to China