AI Agents Pose New Cybersecurity Risks, Experts Warn

Key Takeaways

• Cybersecurity experts warn AI agents need the same identity governance rules as human employees
• Organizations can deploy thousands of agent identities quickly, creating scale and security challenges
• Experts urge zero-trust access, least privilege, and constant monitoring for all AI identities
• Enterprises adopt AI tools faster than they can secure them, according to industry analysis
• Red-teaming, logging, and behavioral AI detection are recommended to reduce AI security gaps

Cybersecurity Awareness Month highlights AI agent risks

Cybersecurity experts say AI agents need the same security rules as human workers. Darren Guccione of Keeper Security says organizations are creating many digital agent identities without proper governance. John Cannava and Michael Marino add that AI agents can access systems and make decisions at machine speed. Poorly managed agents could give attackers access to important data. Security leaders urge zero-trust rules and least-privilege access for all AI identities.

Keeper Security warns AI agents increase cyber risk

Keeper Security says AI agents are creating a new layer of cyber risk by acting like digital users. CEO Darren Guccione says companies often do not apply identity governance to agents the way they do for employees. He warns that scale is a major problem because organizations can deploy thousands of agent identities quickly. Michael Marino says modern Privileged Access Management must go beyond password vaults. Experts urge zero-trust rules, least privilege, and constant monitoring of AI agents.

Four ways to close the AI security gap

This article says enterprises are adopting AI tools faster than they can secure them. It uses the idea that slow and deliberate work leads to faster results. The author calls for least-privilege access, logging, human oversight, and red-teaming before AI systems go live. The article also calls for better visibility into how employees and AI agents use tools and data. Behavioral AI is presented as a way to detect unusual activity faster than traditional rules.

Flow Engineering raises $50M for AI hardware design agents

Flow Engineering raised a $50 million Series B round at a $750 million valuation. The San Francisco startup builds AI tools that help teams design hardware by aligning CAD drawings, requirements, and test data. The round was co-led by Valor Equity Partners and Atreides Management, with Sequoia Capital also participating. Customers include Anduril, Rivian, Joby Aviation, Stoke Space, General Motors PPU, and RV Tech. The money will support safer AI tools for engineering data and expanded features.

Flow Engineering secures funding to speed up hardware design with AI

Flow Engineering says AI can shorten hardware design cycles from months to days. Its platform connects requirements, CAD, simulation, code, and test data into one system. AI agents track design changes and re-check requirements in seconds. Founder Pari Singh says 96% of customers come to Flow on their own. The new funding will expand AI tools, engineering teams, and seek FedRAMP authorization for regulated customers.

Texas Instruments Analog Devices ON Semiconductor ride AI chip demand

Analog chips are essential for AI data centers because they convert electricity into usable data. Texas Instruments saw analog chip revenue grow 29% year over year. Analog Devices reported 40% revenue growth and a 45% stock gain year to date. ON Semiconductor expects AI data center revenue to more than double in 2026. All three companies are well positioned as data center construction increases.

Zscaler ransomware report shows AI powered attacks surging

Zscaler released its 2026 ThreatLabz Ransomware Report showing a 275% year over year increase in attacks. Nearly 900 terabytes of data were stolen. The report highlights how AI is used to target senior executives and exploit tools like Microsoft Teams. This rising threat could increase demand for cloud security platforms like Zscaler.

Analyst says 7 AI stocks are undervalued after sell off

Portfolio manager Adam Johnson believes some AI stocks are now on sale after losing investor enthusiasm. He argues every company will need to become an AI company. He points to Micron Nvidia Onto Innovation Applied Digital CoreWeave Marvell GE Vernova and Sterling Infrastructure. He says these companies now trade at lower multiples than the average S&P 500 company.

Dazzle AI CEO Marissa Mayer explains personal assistant tool

Dazzle AI CEO Marissa Mayer appeared on The Exchange to discuss a new personal AI assistant. The tool analyzes a user's camera roll with permission. It then customizes the AI experience based on that personal data. The assistant is designed to act on the world around the user.

NetApp uses data governance to guide AI storage agents

NetApp says data governance sets the rules for its AI storage agents. Enterprise data is spread across on prem data centers public clouds and neoclouds. NetApp combines a consistent data plane with a unified control plane for agents and humans. The company says humans still own accountability and can override exceptions.

AI code supervision creates new vocabulary gap in CS education

As AI systems take over code generation, computer science education researchers see a shift from writing code to supervising AI. This change reveals a missing vocabulary in the field. The gap in terminology makes it harder to discuss and ensure trustworthy AI delegation.

Xopero Software buys Vigil Guard to expand AI security

Xopero Software acquired Vigil Guard to build a unified AI security and data protection platform. Vigil Guard operates an AI Detection and Response platform that detects prompt injection attacks and prevents data leaks before prompts reach external language models. The move responds to rising AI risks as companies embed sensitive data into language model prompts faster than security procedures can keep up.

New ETF for language model builders tops with Meta

The Global X LLM ETF launched on October 1, 2026, focusing on companies that build large language models. Meta Platforms is the largest holding, and about a third of the fund sits in Hong Kong-based Chinese AI companies like MiniMax and Z.AI. Major developers OpenAI and Anthropic are absent because they are private, and the fund charges a 0.75% fee.

Sources

• Cybersecurity Awareness Month: AI agents are users too, and they need governing like it
• AI agents drive new cyber risk, Keeper Security warns
• Four ways to close the AI security gap
• Flow Engineering Secures $50M Series B at $750M Valuation for AI Hardware Design Agents
• Hardware Design AI Platform Flow Engineering Raises $50M at $750M Valuation
• Analog Chips Are the Silent Winners of the AI Trade. These 3 Growth Stocks Are Capitalizing on the Opportunity.
• Ransomware Report Could Be A Big Moment For Zscaler Stock (ZS)
• 7 AI stocks on sale?: Jon Erlichman
• Dazzle AI CEO Marissa Mayer: We grab your camera roll with your permission and customize the AI to you
• Data governance sets the rules for NetApp's AI storage agents
• Cybernetic and Epistemic: A Missing Vocabulary for Trustworthy Agentic Delegation
• Xopero Software Acquires Vigil Guard
• Wall Street Just Launched an ETF for Large Language Models. Here's What's Inside