AI agents

Pentera's AI Security and Exposure Report 2026 says every surveyed CISO reports AI already in use across their organizations. The article argues that security validation needs AI for adaptive testing, but that fully agentic systems, where AI reasoning governs execution end to end, undermine the repeatability that structured security programs require. It proposes a hybrid model in which deterministic logic defines attack chain execution and AI adapts payloads within that structure.

Swedish vendor Curity announced Access Intelligence, an extension to its Identity Server API IAM platform, to secure autonomous AI agents. Its approach treats agents as a special type of application, issuing OAuth tokens that carry the agent's purpose and intent, with access granted per action at runtime rather than through static permissions.

This AWS blog post explains how to secure AI agents and coding assistants that access AWS resources through the Model Context Protocol (MCP). It argues that agents can do anything their granted entitlements allow, so IAM permissions must be designed as deterministic controls, and it presents three IAM security principles with policy examples.

Cloudflare is making OpenAI frontier models, including GPT-5.4, available to millions of customers within Agent Cloud, a platform for deploying AI agents that handle tasks such as responding to customers, updating systems and generating reports. The Codex harness is now generally available in Cloudflare Sandboxes and will come to Workers AI in the near future.

Microsoft describes the agentic SOC, a security operations model that moves from reacting to incidents toward anticipating attacker movement and reshaping the environment to cut off their paths. The model pairs a threat protection platform with built-in autonomous defense and AI agents that assist analysts with investigation and prioritization. The article is the opening of a series and points to a new whitepaper, The agentic SOC: Your teammate for tomorrow, today.

Microsoft has released the Agent Governance Toolkit, an open-source project that monitors and controls AI agents during execution. It adds a runtime security layer that enforces policies against issues such as prompt injection and maps to OWASP's top 10 risks for agentic systems. The toolkit is in public preview under an MIT license, with components in Python, TypeScript, Rust, Go and .NET.

The source argues that the cybersecurity response to AI-enabled nation-state threats cannot be incremental and must instead be architectural. It is a short opinion piece with no further details on specific attacks, products or figures.

Google DeepMind researchers have mapped a vulnerability category called 'AI Agent Traps'. The source says it lets attackers manipulate, deceive, and exploit AI agents that visit malicious web content.

AWS submitted a response to NIST's Center for AI Standards and Innovation (CAISI) Request for Information on securing agentic AI systems, which act autonomously by connecting to tools and APIs and using LLMs to plan and execute actions at machine speed. The response identifies four foundational security principles and the architectural building blocks that implement them. The source excerpt ends before the remaining principles are described.

Simon Willison was a guest on Lenny Rachitsky's podcast, in an episode released 2 April 2026, discussing agentic engineering. The source highlights his view that GPT 5.1 and Claude Opus 4.5 marked an inflection point in November, after which coding agents mostly do what they are told. He argues that software engineers are a bellwether for other information workers.

Variance has raised $21.5 million for its compliance investigation platform, which is powered by AI agents. The company has raised $26 million in total funding, and the latest investment is intended to fuel platform growth.

SecurityWeek promotes a webinar on what Agentic AI can and cannot solve today. The session also covers real-world breach scenarios tied to disconnected applications.

The article argues that enterprise AI is shifting from chatbots that answer questions to AI agents that reason, plan, and act across enterprise systems autonomously. It states that the security risk of an agent depends on its access to systems and data and on how independently it can act without human approval.

Microsoft's blog post discusses the OWASP Top 10 for Agentic Applications (2026), which outlines risks for autonomous systems that act across workflows using real identities, data access and tools. The post explores the list's key findings and highlights practical mitigations grounded in Agent 365 and foundational capabilities in Microsoft Copilot Studio. Microsoft AI Red Team members helped review the list before publication.

Okta CEO Todd McKinnon discusses how the company is responding to pressure from AI, including the idea that customers could build their own tools instead of paying for SaaS. The interview also covers managing AI agent identity, which McKinnon describes as lying between a person and a system, and the need for a kill switch at the agent level.

Anecdotes co-founder and CEO Yair Kuznitsov argues that enterprise GRC teams hesitate to adopt agentic GRC mainly because of an identity and value question, not technology. He contends that agents can take over evidence gathering, remediation tracking and audit cycles, so practitioners must redefine their role around risk management and the judgment needed to design and oversee agent logic.

AWS describes security principles for deploying agentic AI in financial services, focusing on observability of agentic workflows and fine-grained control over agent tool access permissions. The post presents seven design principles and implementation guidance for meeting regulatory requirements, including SR 11-7 in the US, SS1/23 in the UK, and ECB guidelines in the EU.