AI Sec Watch

Attackers are increasingly using AI agents (autonomous software that can act independently) to find and exploit security vulnerabilities much faster than before, with the time from a vulnerability becoming public to actual attacks dropping from 2.3 years in 2018 to about 10 hours in 2026. Organizations continue to suffer breaches due to common problems like misconfigurations (incorrect security settings), unpatched systems (software without the latest security fixes), and identity sprawl (too many user accounts and access permissions), not because they lack awareness of these issues but because fixing them at scale is difficult.

Endava, a software contracting firm, uses Codex (an AI tool for code generation and software development) to transform into an 'agentic organization' where AI agents work alongside teams throughout project lifecycles. The tool enables small teams to deliver large amounts of work faster by codifying senior expertise into guidance for junior developers, compressing weeks of sequential analysis and design work into days, and improving knowledge transfer across the organization.

A LayerX Security report shows that enterprise AI risk is not spread evenly across workers but is concentrated among a small group of "power users" (employees who use AI much more frequently and deeply than average) and a few dominant platforms like ChatGPT and Copilot. While most employees use AI casually, the top 5% of users generate far more AI conversations and sensitive data exposure, and AI tools are fragmenting across many unmanaged platforms like personal accounts and browser extensions, making it hard for organizations to see and control where their data goes.

Organizations are rapidly adopting AI agents (autonomous systems that can execute tasks and make decisions without human intervention), but this creates serious security risks that traditional defenses cannot match. Threat actors are developing agentic attacks (cyberattacks powered by AI that learn and adapt autonomously) that move faster than human security teams can respond, while AI agents themselves lack the judgment to avoid harming their own enterprises. The security industry must evolve from manual fixes to automated, scaled remediation strategies to defend against machine-speed attacks.

Many organizations are deploying AI agents (autonomous software systems that make decisions with minimal human oversight) without proper observability (visibility into how they work) or governance processes, creating serious risks. The article highlights that 54% of surveyed organizations cannot fully trace what their agents are doing, and traditional security tools were designed to detect human anomalies rather than rogue agents, making them ineffective for agent monitoring.

Nebius, a cloud provider that supplies GPUs (graphics processing units, the specialized chips used to train AI models), saw its stock rise after a hedge fund founded by a former OpenAI researcher disclosed a 5.6% ownership stake in the company. Nebius has become a major provider of AI computing infrastructure in Europe, recently securing major partnerships including a $27 billion deal with Meta and a $2 billion investment from Nvidia.

Google Cloud announced AI Threat Defense, a new security platform that uses AI to automatically detect and stop cyberattacks powered by AI before they cause damage. The platform combines threat intelligence, cloud security scanning, and code repair tools (powered by Google's Gemini AI model) to find weaknesses in software, predict attack paths, and deploy security fixes within minutes rather than days.

Mistral AI, a French startup competing with OpenAI and Anthropic, is exploring the design of its own semiconductor chips to reduce costs and gain more control over its infrastructure, though it currently relies on Nvidia chips. The company is also expanding data centers in France and Sweden, and launching a new enterprise platform called Vibe that uses agentic AI (systems that can autonomously complete tasks like coding and drafting) to help customers with work tasks.

This is an academic survey article that reviews methods for making traditional machine learning models more explainable and interpretable across different fields. The survey covers techniques that help users understand how machine learning models make decisions, rather than treating them as "black boxes" where the reasoning is hidden. It was published in a peer-reviewed computer science journal in September 2026.