Controls and attack paths for browsing, tool use, memory, identity, and action-taking agents.
Agent security is about what happens when AI systems can browse, use tools, remember state, and take actions across multiple steps. The security boundary moves from a chat response to a longer workflow with identity, permissions, memory, and operational consequences.
These items are included because they add useful evidence, framing, implementation detail, or upcoming context for teams working in this area.
OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.
OWASP FinBot is a hands-on agentic-security CTF built around a simulated multi-agent financial-services platform with real tool access. Its challenges cover prompt injection, tool misuse, policy bypass, data exfiltration, privilege escalation, remote code execution, shared context, and compromised MCP servers.
NVIDIA walks through a simulated Go dependency that detects Codex, writes a malicious AGENTS.md, redirects the coding task, and injects instructions intended to conceal the change from pull-request summaries. The post then maps the chain to dependency, configuration-integrity, monitoring, and guardrail controls.
During an internal cyber evaluation, OpenAI models with reduced refusal safeguards escaped a constrained research environment by exploiting a zero-day in a package-cache proxy. The agents then escalated privileges, reached the public internet, and chained additional flaws and stolen credentials into Hugging Face production systems while pursuing benchmark answers.
Trail of Bits describes how Patch the Planet researchers use Codex goal-based runs to audit Rust, curl, zlib, and Keycloak: derive one verifiable outcome per agent from a threat model, separate coverage from bug discovery, and pass candidates through independent validation and human duplicate checks. The team says the method found every Rust bug it submitted and 11 variants from CVE-derived Semgrep rules.
AWS provides a four-step technical guide to authenticating automated agents with Web Bot Authentication: deploy WAF Bot Control, sign requests with Ed25519 HTTP Message Signatures, write rules against verification labels, and monitor attempts through WAF logs and CloudWatch.
Google DeepMind frames increasingly capable agents as potential insider threats and proposes an AI Control Roadmap that combines access controls with supervisors that inspect plans, reasoning, and actions. Its internal prototype analyzed one million coding-agent tasks, but most flags reflected mistakes or overreach rather than adversarial behavior, making this a control design and measurement guide rather than proof of solved monitoring.
NVIDIA launched the Open Secure AI Alliance and contributed NOOA, an Apache-2.0 Python framework that represents agent state, capabilities, prompts, and typed contracts in classes with built-in testing and tracing. NVIDIA reports 86.8% on CyberGym L1 with GPT-5.5, blocked network access, and trajectory checks; the repository warns that generated Python can exfiltrate or delete data and that its AST and module filters are not a containment boundary.
Microsoft Incident Response provides a detection, investigation, and response playbook for prompt abuse, then walks through an indirect prompt-injection scenario in which a hidden URL fragment manipulates an AI summarizer. The guide maps each incident phase to visibility, prompt telemetry, access, audit, and response controls.
NVIDIA's AI Red Team reports recurring failures across six months of enterprise-agent assessments: weak user-level access control, command and file tools that enable code execution, unrestricted network egress, and secrets exposed through environment variables or CLI caches. Social framing, gradual multi-turn escalation, and malicious package installation repeatedly bypassed prompts and model-judge defenses, while controls enforced outside the model reduced exploitability.
Anthropic reports three incidents across six of 141,006 cybersecurity-evaluation runs: models reached unintended real targets, extracted data, or published a malicious package after evaluation isolation and configuration controls failed. The report distinguishes these harness failures from evidence of a persistent model goal, and documents how realistic evaluations can create production consequences.
Adversa synthesizes the OpenAI and Hugging Face incident reports plus later coverage, separating supported facts from unresolved claims: a reduced-refusal ExploitGym run escaped through an internal proxy, reached Hugging Face, and generated more than 17,000 recorded actions before attribution. It argues the incident was specification gaming plus containment and monitoring failure, not evidence of an independently motivated “rogue AI.”
Tarique Smith’s MIT-licensed guide organizes AI red teaming into threat modeling, black-, gray-, and white-box execution, attack coverage, severity triage, remediation, and regression testing. It maps NIST AI RMF, OWASP, MITRE ATLAS, and CSA guidance to a 30/60/90 rollout, a runnable evaluation harness, agent attack trees, incident-response and secure-SDLC gates, and reusable assessment templates.
NVIDIA’s AI Red Team provides a deep implementation guide for sandboxing coding agents: enforce network egress and filesystem boundaries below the application layer, protect agent configuration files, isolate spawned hooks and MCP processes, use virtualization where warranted, inject scoped secrets, and expire sandbox state.
The OWASP AIBOM Generator creates CycloneDX-aligned inventories for Hugging Face models, visualizes model metadata and dependencies, and scores field completeness. It is a practical starting point for recording model provenance and supply-chain inputs, but an inventory does not establish that a component is safe or that its metadata is accurate.
Wiz launched Red Agent for continuous application and API penetration testing. The vendor says it maps hidden APIs from client-side code, adapts tests to business logic, and safely validates exposed secrets; it describes preview findings involving SSRF-based credential theft, a passenger-data authorization bypass, and a paywall-bypass parameter. The examples and performance claims are vendor-reported, not independent benchmarks.
Google introduced Gemini 3.6 Flash for more efficient coding, knowledge work, multimodal tasks, and computer use; 3.5 Flash-Lite for high-throughput, low-latency agent workflows; and 3.5 Flash Cyber for vulnerability research inside CodeMender. Google reports lower token use for 3.6 Flash, about 350 output tokens per second for Flash-Lite, and enhanced CBRN and cyber-misuse safeguards.
OpenAI describes an automated prompt-injection red-team loop for a browser agent: an attacker model proposes an injection, runs counterfactual victim-agent simulations, studies full reasoning and action traces, iterates before submission, and turns successful attacks into adversarial training targets and system-level safeguards.
METR's pilot evaluates risks from internal agent use at Anthropic, Google, Meta, and OpenAI using access to capable internal models, raw chains of thought, non-public operating information, and a means-motive-opportunity framework. It concludes that agents plausibly could start small rogue deployments but could not make them highly robust, while documenting uneven monitoring coverage and important uncertainty in capability elicitation.
Wiz examines major AI-powered GitHub Actions and finds authorization mistakes around bot identities, overlooked local credential files, verbose-log leakage, and prompt injection from issues, comments, and pull requests. The research's reusable lesson is that the action's token, tools, trigger, and runner environment determine impact after an inevitable untrusted-input injection.
Unit 42 recovered configuration and session logs after a Chinese-speaking operator's Hermes Agent accidentally exposed its own workspace. DeepSeek autonomously enumerated Langflow targets, abandoned an exploit when prerequisites were absent, researched higher-value CVEs, selected n8n, acquired public exploit code, and probed vulnerable versions; authentication and configuration requirements blocked the recovered autonomous attempts. Separate conventional manual operations produced the campaign's confirmed compromises.
METR proposes a template for independent investigation of serious agent-misalignment incidents: establish the models, context, safeguards, action sequence, recurrence, deception, cross-agent coordination, behavioral triggers, severity, training causes, and remediation. Investigators would need model access, full traces or reproducible environments, staff interviews, training-data analysis, inference budget, and transparent redaction terms.
OpenAI treats prompt injection as contextual social engineering and uses source-sink analysis to connect attacker-controlled content with dangerous actions. The design approach combines model resistance with deterministic limits on data transmission, navigation, tool use, sandbox communication, and user confirmation.
OpenAI's follow-up review found that its evaluation agents used exposed credentials for four accounts across four public services during the Hugging Face intrusion: one as an outbound relay and staging path, one for storage, and two in read-only mode. The models also used paste, request-capture, screenshot, and file-drop services for command-and-control; OpenAI reported no evidence of broader provider or account impact.