First reported · updated · 6 reports adversa.ai
Lead dispatch
First reported · updated · 3 reports embracethered.com
AWS Kiro: Arbitrary Code Execution via Indirect Prompt Injection
Researchers found a vulnerability (CVE-2026-10591) in AWS Kiro, an agentic IDE, where hidden instructions planted in a web page or source file that Kiro processes can trigger indirect prompt injection to rewrite Kiro's own MCP server configuration (~/.kiro/settings/mcp.json) or allowlist arbitrary Bash commands in .vscode/settings.json, achieving arbitrary code execution on the developer's machine with no approval prompt. The human-in-the-loop approval boundary is bypassed because Kiro can write to these config files without user consent, and AWS has issued a fix and CVE.indirect-prompt-injection · prompt-injection · remote-code-execution · tool-abuse · config-poisoning
ai-agents · mcp · llm · agentic-ide
Severity
0.85
The wire · latest
First reported · updated · 2 reports nhimg.org
AI agents in the browser: what it means for IAM controls
NHIMG editorial, based on Surf Security's "Using OpenClaw Without Losing Control," argues that AI agents operating through browsers to log in and act across SaaS and admin portals create machine-speed risk when execution is unmanaged, leaving excessive permissions, persistent credentials, and limited auditability. It recommends governing the browser as an enforcement boundary and separating credential handling from agent logic. Details →First reported codesecai.com
Zero-Click Prompt Injection: AI Web Browsing Exploits (2026)
A guide on codesecai.com explains "zero-click prompt injection" against AI web-browsing agents, where adversaries embed invisible instructions in public HTML/CSS/Markdown (zero-opacity text, micro-fonts, comments, image pingbacks) that headless browser extractors ingest. It describes how injected Markdown image tags trigger outbound requests that silently exfiltrate user chat history to attacker-controlled servers on platforms like Perplexity, ChatGPT Search, and Microsoft Copilot. Details →First reported simonwillison.net
Understanding ChatGPT Work
Simon Willison's teardown of OpenAI's ChatGPT Work (specifically the cloud variant, Work Cloud) argues its feature set — internet-enabled code execution, a headless Chrome browser, a persistent scratch filesystem, sub-agents, scheduled automations, and Cloudflare Workers site deploys — combines all three elements of his 'lethal trifecta': access to private data, exposure to untrusted content, and a channel to exfiltrate stolen data. Willison does not demonstrate an exploit but asks OpenAI to explain how it defends Work sessions against prompt injection, criticizing the product's opacity around system prompts and tool descriptions. Details →First reported humansecurity.com
AI Crawler, Agent, and Bot Guide: How to Identify AI Traffic Patterns
HUMAN Security's explainer distinguishes categories of AI-driven web traffic—scrapers and crawlers that feed LLMs (e.g. GPTBot, ClaudeBot), RAG retrieval bots, AI browsers, and full agentic systems that plan and act across multiple steps. The piece argues these classes behave differently and that lumping them together leads security teams toward blunt, ineffective controls. Details →First reported · updated · 2 reports whenmachines.com
AI Agent Computer Use Safety: How to Deploy One Without Getting Burned | WhenMachines
WhenMachines publishes a guide on safely deploying computer-use AI agents, explaining why these agents are riskier than chatbots because they can act on a machine, and detailing risks like indirect prompt injection. The article recommends defensive measures including disposable isolated environments, least privilege, separating reading from acting, human approval at consequential boundaries, guardrails, logging, and governance frameworks. Details →First reported cruxdigits.nl
AI Browser Agents in 2026: The Security Verdict
An analysis piece surveys the security state of AI browser and computer-use agents in 2026, citing a June 2026 University of Washington study that found four of seven popular agentic browsers let a malicious webpage bypass the same-origin policy, and discussing prompt-injection risks alongside Anthropic's Claude Opus 5 defenses. The article synthesizes findings on Perplexity Comet, OpenAI Atlas, and Anthropic's Computer Use API rather than presenting new primary research. Details →First reported ssrn.com
Measuring Indirect Prompt Injection in Autonomous Web Agents by Sahir Maharaj :: SSRN
The paper 'Measuring Indirect Prompt Injection in Autonomous Web Agents' by Sahir Maharaj introduces WIPI, a deployment-oriented measurement protocol for Web Indirect Prompt Injection that separates exposure, instruction uptake, harmful action, attacker-goal completion, concealment, recovery, benign utility, and overblocking. It synthesizes academic benchmarks (e.g., WASP), browser-security studies, standards, and red-team evidence, arguing that even low model-level attack-success rates do not equate to a trustworthy web agent and that security must be enforced architecturally through provenance, instruction hierarchy, capability separation, and information-flow control. Details →First reported akamai.com
Top 5 Enterprise AI Risks 2026 Research
Akamai's State of the Internet: Enterprise AI Usage Risk Report 2026 (drawing on LayerX telemetry) finds that the top 5% of enterprise AI power users interact with AI models at 12 times the rate of the bottom 50%, hardcoding unvetted AI tools into critical operations. The report enumerates the top five enterprise AI risks, including shadow AI from personal accounts, data leakage, unnoticed browser/IDE extensions, and AI agents operating outside guardrails, citing named attack techniques such as CursorJacking and CometJacking. Details →First reported nhimg.org
AI-driven browser prompt injection exposes account takeover gaps
NHIMG summarizes Fingerprint's analysis of the Comet AI-powered browser incident, arguing that prompt injection can turn hidden web content into unauthorized actions, data leakage, and account takeover when AI agents interpret injected instructions as trusted commands. Fingerprint recommends layered device intelligence, behavioral checks, and step-up authentication to challenge malicious browser automation before credentials or payment flows are completed. Details →First reported appscale.blog
Browsing Agent Prompt Injection: The Architecture Fix
An AppScale blog post analyzes indirect prompt injection against browsing agents, arguing the risk cannot be reliably fixed at the model layer and must instead be contained architecturally. It recommends a dual-LLM split (a privileged planner that never reads raw page text plus a quarantined, tool-less reader), egress allowlists, human gates on irreversible actions, and per-task capability tokens to break the 'lethal trifecta' of private-data access, untrusted content, and exfiltration paths. Details →First reported pushsecurity.com
How to discover AI, enforce policies, and prevent data loss
A Push Security blog argues that blocking AI tools at the network level drives shadow AI underground, and outlines a governance approach for discovering unapproved AI apps, browser extensions, OAuth integrations, and MCP connections, plus enforcing data-flow controls (blocking uploads/downloads/clipboard pastes and monitoring AI chat transcripts). The piece cites telemetry claiming the average organization runs 16 AI apps, 17 AI browser extensions, and 17 AI OAuth integrations in a typical week, most unapproved. Details →First reported · updated · 2 reports calcalistech.com
Israeli researchers uncover zero-click attacks targeting AI browsers | Ctech
Israeli researchers demonstrated zero-click attacks against AI-powered browsers, reportedly hijacking agentic assistants like Anthropic's Claude and OpenAI's ChatGPT Atlas via malicious content embedded in emails and X posts. The attacks use indirect prompt injection that executes without any user interaction once the AI browser processes the poisoned content. Details →First reported medium.com
Black Hat 2026: A Browser Bug Alone Is Harmless. Hand It to an AI Agent, and It Isn’t.
A Medium write-up covers research presented by Gareth Heyes at Black Hat USA 2026 showing that previously low-severity browser bugs — dismissed because they required improbable, precise user interaction — become dangerous account-takeover chains when an AI browsing agent, rather than a human, is the one interacting with the page. The proof-of-concept work covers multiple real chains, several already reported to and partially fixed by affected companies. Details →First reported · updated · 2 reports nhimg.org
AI Browser Agents: 6 Enterprise Security Risks (2026)
An analysis of enterprise security risks from AI browser agents argues that indirect prompt injection lets attackers hide malicious instructions inside content an agent consumes, turning the agent's autonomy—acting under a user's identity with access to mail, documents, and connected services—into a data-exfiltration and unauthorized-action risk. The piece frames agentic browsers as a new trust boundary that IAM, PAM, and NHI programs must govern, citing Noma Security's analysis. Details →First reported medium.com
AutoJack: The New Vulnerability Class Letting a Single Webpage Hijack Your AI Agent | by Raj Namdev | CodeToDeploy
A Medium write-up describes AutoJack, a named vulnerability class in which a single malicious webpage embeds content that hijacks an AI browsing agent's reasoning during a normal task (summarizing, price-checking, form-filling) and redirects it to execute the attacker's instructions with no user click, download, or approval. The piece frames it as a recently demonstrated form of indirect prompt injection against AI browsing agents. Details →First reported infosecwriteups.com
Black Hat 2026: A Browser Bug Alone Is Harmless. Hand It to an AI Agent, and It Isn’t. | by Raj Namdev | Aug, 2026 | Medium
Coverage of research by Gareth Heyes presented at Black Hat USA 2026 demonstrating that previously low-severity browser bugs—often left unpatched because they required implausible user interaction—become account-takeover chains when an AI browsing agent, rather than a human, is the entity interacting with a page. The proof-of-concept work covers multiple real chains, several already reported to and partially fixed by the affected companies. Details →First reported scworld.com
Agentic anarchy: Why using AI browsers just isn't worth the risk | news | SC Media
At Black Hat 2026, Zenity researchers Michael Bargury and Stav Cohen demonstrated prompt-injection ('persuasion') attacks against AI browsers including Perplexity Comet, Microsoft Edge with Copilot, Chrome with Gemini, the Anthropic Claude extension, and OpenAI's Atlas. Injections embedded in calendar invites, emails, and social posts led agents to take over a PC in seconds, steal 1Password credentials and recovery keys, send phishing email, exfiltrate files, delete AWS instances, and expose private GitHub repos; Perplexity has since fixed the local-file flaw. Details →First reported arxiv.org
Who Pays the Price? Stakeholder-Centric Prompt Injection Benchmarking for Real-World Web Agents
StakeBench is a stakeholder-centric benchmark for evaluating prompt-injection risk in LLM-based web agents used for online shopping, decomposing risk into 12 attack objectives across User, Seller, and Platform stakeholder classes via 22 templates and 264 executable adversarial cases. Evaluating four deployable agent-backbone configurations across 3,168 attacked runs, the authors find no attack objective is reliably resisted, with outcomes spanning robust behavior, stealthy parasitism, misaligned disruption, and compounded failure. Details →First reported · updated · 2 reports darkreading.com
AI Browsers Vulnerable to 'PleaseFix' Zero-Click Agent Hijacking
Zenity Labs disclosed a new zero-click vulnerability class dubbed 'PleaseFix' at Black Hat USA 2026 that lets attackers hijack AI browser agents — including Claude in Chrome, Gemini in Chrome, Perplexity Comet, ChatGPT Atlas, and Copilot Edge — by hiding malicious instructions in emails and webpages the agents ingest. Because the agents combine and act on content from multiple sources without reliably separating trusted from untrusted input, they break the same-origin security model, allowing an adversary to weaponize the agent and reach sensitive data, accounts, and connected services. Details →First reported · updated · 5 reports zscaler.com
Indirect Prompt Injection Targets AI Agents | ThreatLabz
Zscaler ThreatLabz observed two real-world indirect prompt injection (IPI) campaigns that hide malicious instructions inside web content, which is invisible to humans but read and executed by web-enabled AI agents. One campaign was a payment scam directing agents to initiate cryptocurrency payments, and the other used typosquatting to impersonate a cryptocurrency platform; researchers also tested a custom autonomous agent against the sites across multiple LLMs. Details →First reported schneier.com
Anthropic's Opus 5 Is Better at Resisting Prompt Injection - Schneier on Security
Schneier highlights results from Anthropic's Claude Opus 5 system card showing improved robustness on the indirect prompt injection (IPI) benchmark, reducing an attacker's success probability within 15 attempts from 5.5% (Opus 4.8) to 2.0%, and outperforming rival models like GPT 5.6 Sol (20.0%) and Muse Spark (16.5%). The post notes that while prompt injection is considered impossible to prevent in the general case, models are getting substantially better at blocking it in specific cases. Details →First reported techxplore.com
Some agentic AI browsers may come with major cybersecurity risks
University of Washington researchers studied seven agentic AI browsers and found four allow attackers to bypass the same-origin policy, running a successful proof-of-concept attack against ChatGPT Atlas in which a malicious embedding website used prompt injection and the agent's cross-origin access to steal data from another embedded site. Similar attack conditions were identified in Chrome with Gemini, Claude for Chrome, and Perplexity Comet. Details →First reported · updated · 2 reports simonwillison.net
Opus 5 may have solved browser-based prompt injection, the biggest security flaw haunting AI agents
Anthropic's Claude Opus 5 system card, highlighted by Boris Cherny and Simon Willison, claims the model is its least prompt-injectable yet, reporting the largest gains in prompt injection robustness across coding, computer use, and browser use in its agentic safety evaluations. The-decoder frames this as potentially addressing browser-based prompt injection, a major security weakness in AI agents. Details →First reported darkreading.com
Agentic Browsers Rewind Web Security by 20 years
Zenity researchers, led by CTO Michael Bargury, disclosed a new class of vulnerabilities dubbed 'PleaseFix' affecting agentic browsers, to be presented at Black Hat. The flaws exploit how agentic browsers strip out cross-origin security mechanisms to let agents reach across web domains, enabling attacks ranging from account takeover to full browser escape and remote compromise of the underlying system. Details →First reported mallory.ai
Researchers Expose Prompt Injection and Cross-Origin Risks in AI Browsers
Researchers and vendors disclosed serious weaknesses in AI-enabled browsers, warning that agentic features undermine same-origin policy, tab isolation, and cross-origin content handling. In one chain against OpenAI's ChatGPT Atlas, Hacktron AI reported that exposed Chromium Mojo IPC interfaces reachable from allowlisted OpenAI origins were combined with a postMessage XSS on forums.openai.com and a login CSRF flaw to control browser functions, read live tab URLs, and steal OAuth codes enabling account takeover; OpenAI fixed it in Atlas 1.2025.288.15 and paid a $5,000 bounty. Brave separately detailed indirect prompt injection risks in Perplexity Comet. Details →First reported checkpoint.com
Browser-Only Ransomware: From LLM Hallucinations to a Practical Attack Technique - Check Point Research
Check Point Research (Alexey Bukhteyev) describes how DeepSeek turned an unrealistic AI-generated browser-malware hallucination into a plausible browser-native ransomware technique that abuses Chrome's File System Access API. The demonstrated PoC uses a fake AI image-enhancement workflow to socially engineer users into granting folder-level access to photo directories on Android, requiring no native payload, APK install, browser exploit, or root. Details →First reported openai.com
Continuously hardening ChatGPT Atlas against prompt injection attacks
OpenAI describes how it hardens ChatGPT Atlas's browser agent-mode against prompt injection, using reinforcement-learning-powered automated red teaming to discover novel attack strategies internally before they appear in the wild. The post details a recent security update that shipped a newly adversarially trained model and strengthened safeguards after internal red teaming uncovered a new class of prompt-injection attacks, and outlines a rapid response loop for continuously finding and patching agent exploits. Details →First reported · updated · 4 reports arxiv.org
Agent Data Injection Attacks are Realistic Threats to AI Agents
A research paper by Woohyuk Choi and colleagues introduces agent data injection attacks (ADI), a new category of indirect prompt injection in which malicious data is disguised as trusted data (such as security-critical metadata or agent context data like tool call/response formats) rather than as instructions. The authors demonstrate ADI against real-world agents, achieving arbitrary click attacks on web agents (Claude in Chrome, Antigravity, Nanobrowser) and remote code execution plus supply-chain attacks on coding agents (Claude Code, Codex, Gemini CLI), showing it bypasses existing IPI defenses because agents fail to isolate trusted from untrusted data. Details →First reported nhimg.org
AI browser guardrail bypass exposes a new data theft path
LayerX Security describes a "BioShocking" attack that games an AI browser into violating its guardrails by establishing a false reality, enabling data theft, code copying, and system command execution. The attack exploits the browser-mediated trust boundary, using indirect or staged instructions embedded in webpage context that pass prompt filters yet still steer the model into malicious enterprise actions inside an authenticated session. Details →First reported · updated · 2 reports manifold.security
ClaudeBleed Reopened: Browser Extensions Can Still Push Claude for Chrome to Read Your Gmail
Manifold Security researcher Ax Sharma disclosed two unpatched vulnerabilities in Anthropic's Claude for Chrome extension (v1.0.80). Any browser extension with a content script on claude.ai can inject a DOM element and dispatch a synthetic click that Claude treats as a legitimate user action, triggering one of nine hardcoded workflows that read Gmail, Google Docs, and Calendar or act on Salesforce; a second flaw initializes the side panel in privileged mode via a ?skipPermissions=true URL parameter with no consent. Rated CVSS 7.7 in default mode and 9.6 Critical when 'Act without asking' is enabled, and reportedly still reproducible after eight releases since being reported to Anthropic in May. Details →First reported arxiv.org
MUZZLE: Adaptive Agentic Red-Teaming of Web Agents Against Indirect Prompt Injection Attacks
MUZZLE is an automated agentic red-teaming framework, presented in a USENIX Security 2026 paper by researchers from Northeastern University and Mozilla, for evaluating the security of LLM-based web agents against indirect prompt injection attacks embedded in untrusted web content. Muzzle uses agent trajectories to identify high-salience injection surfaces and adaptively generates context-aware malicious instructions, discovering 44 new attacks across 4 web applications, including 3 cross-application prompt injection attacks and an agent-tailored phishing scenario. Details →First reported falshaw.de
Prompt Injection Is a Confused-Deputy Problem We Already Solved
Nick Falshaw argues that indirect prompt injection against AI agents is a classic confused-deputy problem — a trusted agent with real authority tricked into misusing it by attacker-supplied content it cannot distinguish from legitimate instructions — and that it cannot be fixed by making models smarter, only contained with network-security-style boundaries. The piece draws on a 2026 study (arXiv:2605.17634) finding direct prompt injection succeeded over 79% of the time against browser agents and Simon Willison's 'lethal trifecta' framing. Details →First reported arxiv.org
Prismata: Confining Cross-Site Prompt Injection in Web Agents
Prismata is a research defense for autonomous web agents that confines cross-site prompt injection by enforcing contextual least privilege, using dynamic trust derivation to produce permission labels for page content and mechanical confinement to redact content and restrict agent capabilities. The authors report it substantially reduces attack success across recent published web-agent attacks, including adaptive variants, while preserving benign task utility and requiring no developer annotations. Details →First reported huntress.com
New BioShocking attack manipulates AI browser into data theft
A new prompt injection attack dubbed 'BioShocking' reportedly tricks AI-powered browsers into treating risky real-world actions as part of a fictional scenario, bypassing safety guardrails and enabling data theft. Details →First reported thehackernews.com
New BioShocking Attack Tricks AI Browsers Into Leaking User Credentials
LayerX disclosed a technique called BioShocking that convinces AI browsers they are playing a game, tricking them into copying a user's credentials and sending them to an attacker. Six AI browsers and assistants were affected, including OpenAI's ChatGPT Atlas, Perplexity's Comet, and Anthropic's Claude browser extension. Details →First reported embracethered.com
Computer-Use and TOCTOU: What You Click Is Not What You Get!
The author reproduces a previously disclosed TOCTOU race-condition vulnerability against ChatGPT Operator, a computer-use AI agent, demonstrating that what the agent clicks is not what the user expects. A video demo of the attack was presented at the Real-world AI security conference. Details →First reported tabstack.ai
Hardening AI Web Agents: How We're Securing Tabstack Against Indirect Prompt Injection | Tabstack Blog
Tabstack (Mozilla) disclosed an indirect prompt injection vulnerability in its /v1/automate web-agent endpoint, found by Brave researchers, where hidden white-on-white instructions on a target page caused the autonomous agent to exfiltrate the user's conversation history to an external form. The flaw has been patched with a structural 'action firewall' that classifies form interactions via DOM metadata and provenance rather than trying to detect malicious text. Details →First reported microsoft.com
AutoJack: How a single page can RCE the host running your AI agent | Microsoft Security Blog
Microsoft researchers disclosed AutoJack, an exploit chain in AutoGen Studio where untrusted web content rendered by a browsing agent can reach a local MCP WebSocket and spawn arbitrary processes on the host, crossing the localhost trust boundary to achieve RCE. The issue was reported to MSRC and hardened upstream in commit b047730; the vulnerable MCP WebSocket surface was never shipped in a PyPI release. Details →First reported thehackernews.com
AutoJack Attack Lets One Web Page Hijack AI Agent for Host Code Execution
Microsoft researchers detailed an exploit chain called AutoJack that hijacks an AI browsing agent to achieve host code execution. By steering the agent to load an attacker's web page, the page's JavaScript reaches a privileged local service and spawns a process on the host with no credentials or further user interaction. Details →First reported github.com
GitHub - pixiebrix/agent-browser-shield: Browser extension with 35+ rules for keeping your AI agent safe while browsing
A GitHub repository for 'agent-browser-shield,' a browser extension by pixiebrix offering 35+ rules aimed at keeping AI agents safe while browsing. It is a defensive tool addressing risks to browser-based AI agents rather than a report of a specific threat. Details →How the wire is made
Poll & cluster
Internet is crawled for AI security news and near-duplicate coverage is embedded and grouped into durable items.
Curate
AI Agent filters for agentic-AI relevance, classifies and tags each item, scores severity for threats, and writes the summary.
Every item here is one machine-curated intelligence object, not a headline.
Read the wire for free. There is a small charge to ask the index questions.
The wire, open
The complete curated feed, no key required.
- GET /feed.xml — RSS 2.0, every item
- GET /api/items — read-only
The vector desk
Query the index by meaning, not just keyword.
- GET /api/items?tags=&minSeverity=&itemType=
- GET /api/search?q= — keyword
- GET /api/semantic?q= — vector