PT-2026-83643

Name of the Vulnerable Software and Affected Versions ash ai versions 0.1.0 through 0.9.x Description An improper control of code generation allows a remote, unauthenticated client to execute arbitrary Elixir code. The AshAi.Actions.Prompt function evaluates prompt content using EEx.eval string/2...

PT-2026-83668

Loop with Unreachable Exit Condition Infinite Loop vulnerability in ash-project ash ai allows an attacker who can influence a model's output to hang the tool loop and drive unbounded, repeated model requests. AshAi.ToolLoop classifies a model response of :tool calls, then filters the calls throug...

The Fragility of Jailbreak Robustness across Operational States

Existing jailbreak evaluations typically characterize robustness using a single attack success rate ASR measured in a default configuration the vanilla state. However, user-LLM interactions can induce diverse operational states beyond the vanilla state. In this work, we find that jailbreak...

PT-2026-83818

Name of the Vulnerable Software and Affected Versions Eclipse Theia versions 1.73.0 through 1.74.x Description In AI Agent Mode, file-change tools resolve model-supplied file paths without verifying if they are contained within the workspace. This allows a crafted relative path, an absolute path,...

Beyond the Payload: How User Invocation Shapes Coding Agent Vulnerability to Repository Poisoning

Coding agents are increasingly used for software engineering tasks, including bootstrapping projects from third-party repositories whose integrity cannot be assumed. Prior work on repository poisoning largely focuses on attacker-controlled injection and disguise, but developers also shape risk...

ECLIPSE: Self-Evolving Stealthy Prompt Injection Attack against Long-Horizon Agentic Systems

Recently, large language model LLM agents, such as Codex, Claude Code, and OpenClaw, have become capable of planning and executing long-horizon tasks through repeated tool calls. This capability also creates new opportunities for prompt injection. Existing attacks either place the malicious...

SIR: Self-Improving Red-Teaming for Compute Use Agents

Computer use agents CUAs are vision-language models that perceive a screen and act on a real operating system through mouse, keyboard, and terminal, and they are increasingly deployed to automate everyday digital tasks. Because they can be exposed to untrusted content while operating, they are...

CVE-2026-82280

Quivr through 0.0.322 fails to validate ownership in prompt endpoints, allowing authenticated users to modify any prompt by identifier. Attackers with read-only access to shared brains can read exposed prompt identifiers and overwrite system prompts affecting all brain users...

CVE-2026-70331

Improper neutralization of input used for llm prompting in Microsoft Edge for iOS allows an unauthorized attacker to perform spoofing over a network...

EUVD-2026-67846

Improper neutralization of input used for llm prompting in Microsoft Edge for iOS allows an unauthorized attacker to perform spoofing over a network...

GHSA-2WVM-8MVP-22QV Pocket-ID has an Open Redirect on the OIDC /authorize page via unvalidated redirect_uri with prompt=none

Summary The OIDC authorization page in the pocket-id frontend redirects the browser to an attacker-controlled URL without consulting the backend redirecturi allow-list when the request uses prompt=none. An attacker who knows a valid clientid can craft an /authorize link that sends a victim or a...

Pocket-ID has an Open Redirect on the OIDC /authorize page via unvalidated redirect_uri with prompt=none

Summary The OIDC authorization page in the pocket-id frontend redirects the browser to an attacker-controlled URL without consulting the backend redirecturi allow-list when the request uses prompt=none. An attacker who knows a valid clientid can craft an /authorize link that sends a victim or a...

CVE-2026-55834 Pocket ID: Open Redirect on the OIDC /authorize page via unvalidated redirect_uri with prompt=none

Pocket ID is an OIDC provider that allows users to authenticate with their passkeys to services. From 2.6.0 until 2.9.0, frontend/src/routes/authorize/+page.ts reads the redirecturi query parameter and frontend/src/routes/authorize/+page.svelte uses the raw callbackURL in redirectWithError when...

CVE-2026-55834 Pocket ID: Open Redirect on the OIDC /authorize page via unvalidated redirect_uri with prompt=none

Pocket ID is an OIDC provider that allows users to authenticate with their passkeys to services. From 2.6.0 until 2.9.0, frontend/src/routes/authorize/+page.ts reads the redirecturi query parameter and frontend/src/routes/authorize/+page.svelte uses the raw callbackURL in redirectWithError when...

EUVD-2026-67828

Pocket ID is an OIDC provider that allows users to authenticate with their passkeys to services. From 2.6.0 until 2.9.0, frontend/src/routes/authorize/+page.ts reads the redirecturi query parameter and frontend/src/routes/authorize/+page.svelte uses the raw callbackURL in redirectWithError when...

CVE-2026-55834 Pocket ID: Open Redirect on the OIDC /authorize page via unvalidated redirect_uri with prompt=none

Pocket ID is an OIDC provider that allows users to authenticate with their passkeys to services. From 2.6.0 until 2.9.0, frontend/src/routes/authorize/+page.ts reads the redirecturi query parameter and frontend/src/routes/authorize/+page.svelte uses the raw callbackURL in redirectWithError when...

CVE-2026-82280

Quivr through 0.0.322 fails to validate ownership in its prompt endpoints, allowing any authenticated user to modify prompts by identifier. An attacker with only read-only access to a shared brain can discover exposed prompt identifiers and overwrite system prompts , affecting all users of that b...

CVE-2026-82280 Quivr Prompt Endpoints Missing Ownership Validation

Quivr through 0.0.322 fails to validate ownership in prompt endpoints, allowing authenticated users to modify any prompt by identifier. Attackers with read-only access to shared brains can read exposed prompt identifiers and overwrite system prompts affecting all brain users...

EUVD-2026-67811

Quivr through 0.0.322 fails to validate ownership in prompt endpoints, allowing authenticated users to modify any prompt by identifier. Attackers with read-only access to shared brains can read exposed prompt identifiers and overwrite system prompts affecting all brain users...

CVE-2026-82280 Quivr Prompt Endpoints Missing Ownership Validation

Quivr through 0.0.322 fails to validate ownership in prompt endpoints, allowing authenticated users to modify any prompt by identifier. Attackers with read-only access to shared brains can read exposed prompt identifiers and overwrite system prompts affecting all brain users...