UTC

AutoGPT Platform

By Significant Gravitas. 111 published vulnerabilities.

Advisories
111
Critical
13
Exploited in the wild
None listed
Latest
2026-11-17
Vulnerabilities
SeverityVulnerabilityPublishedFixed in
MediumAutoPilot's local sandbox buffers unbounded stdout and stderrGHSA-xm4r-96wv-34m52026-11-17See advisory
MediumAny signed-up user can make the platform pay for unlimited Ideogram/Replicate agent-thumbnail image generations, with no charge, balance check or per-user limitGHSA-x4vp-rcwv-9hx42026-11-17See advisory
MediumCredit top-up refunds and lost disputes leave credits spendable until the webhook, and treat Stripe's cents as creditsGHSA-wwfj-25jr-ffg42026-11-17See advisory
MediumAnyone with a chat share link can read secrets in tool-call arguments and large tool results that skip redactionGHSA-wmfr-wj7p-4v992026-11-17See advisory
MediumA captured Exa Websets webhook delivery can be replayed at any time to re-run the owner's triggered workflowGHSA-wfr2-333w-xrqc2026-11-17See advisory
MediumAutoPilot credit check race: run_capability can run a paid block many times, or for free, for the price of oneGHSA-v77v-f222-gp6g2026-11-17See advisory
HighUnbounded request-validation errors stall the REST API and write rejected request bodies to the logGHSA-v39w-3j4r-c33g2026-11-17See advisory
MediumOpen Redirect via Unvalidated redirect_uri in OAuth Authorize PageGHSA-rr95-66gw-xm5f2026-11-17See advisory
MediumDNS-rebinding TOCTOU in Ollama LLM provider endpoint validation allows SSRF to internal servicesGHSA-rmj7-pw6w-hf332026-11-17See advisory
HighAutoGPT Platform Backend: MCPToolBlock Auto-Lookup Lets a Shared Agent's Author Use the Runner's MCP CredentialsGHSA-rjgp-whm3-w5mj2026-11-17See advisory
MediumRegular Expression Denial of Service (ReDoS) in GitHub CI Results Block (search_pattern)GHSA-rcc6-v5j6-rw4j2026-11-17See advisory
MediumFailed or stopped LLM block calls can refund the full credit charge after the provider already received the requestGHSA-r9jq-jjw9-49j62026-11-17See advisory
MediumOpen Redirect in OAuth authorize endpoint via unvalidated error redirectsGHSA-r87r-9p73-r6v62026-11-17See advisory
MediumAn approved marketplace listing can run its installer's own workflows as the installer, with the author's inputsGHSA-r833-g67j-55gc2026-11-17See advisory
HighOAuth Redirect URI Leads to JavaScript ExecutionGHSA-r326-2hff-vh8g2026-11-17See advisory
MediumPublic webhook endpoints buffer large unauthenticated bodies before verificationGHSA-r2xf-4rwf-26r72026-11-17See advisory
MediumOffice preview expands an unbounded embedded ZIP thumbnailGHSA-qxw2-mr3h-jf6g2026-11-17See advisory
MediumHost-scoped credential headers leaked to other hosts through HTTP redirectsGHSA-qx75-7566-5phx2026-11-17See advisory
MediumBrowser SSRF via unguarded redirects, subsequent navigation and DNS re-resolutionGHSA-q5jq-5368-v5cf2026-11-17See advisory
MediumSendEmailBlock SMTP DNS rebinding bypasses SSRF blocklistGHSA-px96-3j32-c68c2026-11-17See advisory
HighAnyone with an Exa Search Monitor trigger's URL can start its workflow with fake events, even with a secret setGHSA-pc98-63jj-97gm2026-11-17See advisory
MediumA workflow that runs itself as a sub-workflow fills every executor worker, blocking other users' runsGHSA-p55h-7x25-pp6g2026-11-17See advisory
CriticalA user can add a version under another user's workflow ID and take over its Marketplace listingGHSA-mwr6-g4xq-4rvp2026-11-17See advisory
MediumAutoPilot never cleans up the media files its block runs write to the executor's disk, so a user can fill itGHSA-jqh7-c84w-36vf2026-11-17See advisory
MediumAny member of a Slack workspace, Telegram group or Teams team can link it to their own AutoGPT accountGHSA-hwcj-243r-mp6x2026-11-17See advisory
MediumCredentials followed a redirect off-origin when the Location header hid the new host behind a backslashGHSA-hp9v-vh5x-5jv32026-11-17See advisory
LowTOCTOU Race Condition in Refresh Token RotationGHSA-h67f-cg97-pj372026-11-17See advisory
CriticalAny signed-up user can reach other users' AgentMail inboxes and Exa Websets through the platform's shared vendor keysGHSA-gfrp-hrh7-x8x82026-11-17See advisory
MediumOffboarded members can still execute org-billed presets and retained webhooks after org removalGHSA-g7j2-3x5v-7x7v2026-11-17See advisory
MediumTemplate rendering was not bounded in size or timeGHSA-f4f5-9hp8-cprw2026-11-17See advisory
MediumUnbounded response buffering in the HTTP helper: memory-exhaustion DoS on URL downloadsGHSA-crrx-44jx-c4q82026-11-17See advisory
HighA Slack workspace member can make the Slack bot send its bot token to their server and fetch internal URLsGHSA-cc2m-58f6-pg2c2026-11-17See advisory
MediumSandbox file extraction has no cumulative byte or file-count limitGHSA-98g6-r386-569h2026-11-17See advisory
MediumPresets and triggers on another user's workflow ran and were billed in that workflow owner's organizationGHSA-8x6w-2c8v-c57j2026-11-17See advisory
HighAn EXECUTE_BLOCK-only API key or OAuth token could run AutoPilotBlock through the external API and get a full, unattended AutoPilot turn with the key owner's stored credentials, agents and workspaceGHSA-8vm3-rxfw-58672026-11-17See advisory
MediumAutoMod timeouts bypass moderation when automodfailopen=FalseGHSA-8mgj-2rcx-pwm42026-11-17See advisory
HighDenial of service and SSRF via a JSON Schema nested in MCPToolBlock and AgentExecutorBlock inputsGHSA-8c79-3c9x-p2v62026-11-17See advisory
MediumAutoPilot tool calls, dry runs and chat-bot turns spend/run past the platform's own usage, concurrency and duration capsGHSA-8752-6cg5-wf782026-11-17See advisory
MediumOpen redirect in the OAuth callback route: the "next" parameter was not sanitized (incomplete fix of GHSA-j2cp-jg5q-38wj)GHSA-7pv2-q2ff-2wx62026-11-17See advisory
MediumSSRF blocklist gap: IPv6 transition prefixes (6to4 / Teredo / NAT64 / IPv4-translated) reach embedded IPv4 endpoints via Requests wrapperGHSA-7334-956w-67pg2026-11-17See advisory
MediumAutoGPT Platform E2B code-execution blocks leave sandboxes running on the operator's shared credential far past what the block actually charges forGHSA-6jc8-g72m-gf952026-11-17See advisory
MediumDenial of service via unbounded format Jinja filter width in TextFormatter (un-patched sibling of GHSA-ppw9)GHSA-5238-wm28-5fvf2026-11-17See advisory
MediumLLM and other dynamically-priced blocks run free when a priced input is left at its schema defaultGHSA-4jw5-mmcx-c2rr2026-11-17See advisory
HighPersonal-organization conversion race lets users bill agent runs to another user's wallet or evade LLM usage chargesGHSA-47r2-jjh5-pcgh2026-11-17See advisory
MediumA member removed from an organization keeps creating workflows in it with an API key made while a memberGHSA-458m-58gm-8cr92026-11-17See advisory
MediumAutoPilot's approval gate and workflow review pause can be bypassed via resume, sub-session delegation, and run_agent version driftGHSA-389q-9q6m-8pxm2026-11-17See advisory
MediumMarketplace and shared agents can redirect a user's saved SMTP or database login to a server the agent's author controlsGHSA-35gg-95rx-c8fm2026-11-17See advisory
MediumOne user can stall other users' workflow runs by leaving executor workers stuck in a video encode or email sendGHSA-2jw6-9qj6-7pr92026-11-17See advisory
MediumBilling-manager-only members are treated as ordinary members and can read other team members' runsCVE-2026-129984 · GHSA-6f2g-95mv-mgwv2026-11-16See advisory
MediumAnyone with an active OAuth app's client ID can stall the REST API by sending wrong client secretsCVE-2026-129317 · GHSA-fp4g-p7j5-v6pw2026-11-16See advisory
MediumA crafted CSV file referenced in AutoPilot can freeze the process running the chatGHSA-q2g2-67rp-2c5r2026-11-12See advisory
HighExposure of Sensitive Information to an Unauthorized Actor and Improper Enforcement of Behavioral Workflow in AutoGPTGHSA-xfwx-rhmm-rjqc2026-10-29See advisory
MediumPublic chat share leaks secrets stored under camelCase keys (redaction bypass)GHSA-x9c2-wpmm-2mx42026-10-29See advisory
MediumReDoS in MatchTextPatternBlock via unbounded user-supplied regexGHSA-qrmh-27g7-9cp22026-10-28See advisory
MediumQuadratic-time ReDoS in XMLParserBlock via the gravitasml tokenizerGHSA-847j-hphx-v6f32026-10-28See advisory
LowConcurrent chat file download can restore access after AutoGPT share revocationGHSA-qv8m-h6gh-rj662026-10-27See advisory
MediumBilling vulnerability: failed nested tool calls refund credits that were never chargedGHSA-p835-fqv4-4x3m2026-10-27See advisory
HighAutoPilot: chat text was pre-processed by the Claude Code CLI (slash commands and file mentions)GHSA-hf7q-6hwp-f6v22026-10-27See advisory
MediumBilling Vulnerability: AutoPilot runs paid blocks on the operator's credential without charging creditsGHSA-f3mw-f2qc-wh9m2026-10-27See advisory
HighSSRF: the unspecified address :: bypasses the request blocklist and reaches loopback servicesGHSA-3c7g-v5hv-wwv42026-10-27See advisory
MediumThe Sentry tunnel forwards first-party cookies, including the session token, to Sentry's ingestGHSA-2739-q98r-hx722026-10-27See advisory
HighHardcoded Default Fernet Encryption Key Exposes All Stored OAuth CredentialsGHSA-57mf-wqwq-6g6x2026-10-05See advisory
CriticalHardcoded Default JWT Secret Enables Complete Authentication BypassGHSA-24q6-6h89-f9p72026-10-05See advisory
HighWebhook provider path confusion bypasses generic webhook secret verificationCVE-2026-72922 · GHSA-349p-3c3r-8mjr2026-08-05See advisory
MediumPreset creation can bind a foreign webhook, enabling known-ID forged victim webhook executions and exposing its signing secretGHSA-4m2w-qfr5-9f3v2026-07-21See advisory
MediumAutoGPT Classic: HTTP client domain allowlist bypass via userinfo and backslash URL tricksGHSA-xxpr-ccx5-48mf2026-07-14See advisory
HighAutoGPT Classic: SSRF in llamafile setup script via unvalidated download URL and redirectsGHSA-vm4v-5hgj-rq772026-07-14See advisory
HighAutoGPT Classic: SSRF in web fetch commands via missing internal/private address validationGHSA-vj3m-g4cv-8j932026-07-14See advisory
HighAutoGPT Classic: GitHub credentials sent to arbitrary hosts via clone_repository URLGHSA-c543-3829-f27q2026-07-14See advisory
LowAutoGPT Classic: litellm version constraint permitted compromised releases (supply chain)GHSA-4hqw-7f5j-5v662026-07-14See advisory
HighAutoGPT Classic: incomplete URL validation in forge allows SSRF in browser commandsGHSA-3cwv-4w3r-xqff2026-07-14See advisory
MediumIDOR in Webhook Ping Endpoint Allows Enumeration and Cross-User Ping TriggeringCVE-2026-56823 · GHSA-rq9m-xvc7-v9h62026-06-26See advisory
HighSSRF in SendWebRequestBlock via IPv4-mapped IPv6 / CGNAT IP-validation bypassCVE-2026-56663 · GHSA-8qc5-rhmg-r6r62026-06-24See advisory
HighDenial of Service (DoS) via Resource Exhaustion in text templating featuresCVE-2026-33235 · GHSA-ppw9-h7rv-gwq92026-06-23See advisory
MediumUser-controlled regex in ExtractTextInformationBlock produces O(n^2) output, enabling memory-exhaustion DoSCVE-2025-32423 · GHSA-pppq-xx2w-7jpq2026-06-23See advisory
MediumQuadratic memory amplification in AITextSummarizerBlock chunking (DoS via attacker-controlled max_tokens/overlap)CVE-2025-32394 · GHSA-955p-gpfx-r66j2026-06-23See advisory
HighDOM-Based XSS and Open Redirect in AutoGPT SignUp PageCVE-2026-55237 · GHSA-j2cp-jg5q-38wj2026-06-17See advisory
CriticalThere is a DoS vulnerability in MediaDurationBlockCVE-2025-32437 · GHSA-rg6v-m9x9-7wf92026-06-11See advisory
HighThere is a DoS vulnerability in AddAudioToVideoBlockCVE-2025-32436 · GHSA-g26x-xwc5-7p442026-06-11See advisory
CriticalThere is a DoS vulnerability in ScreenshotWebPageBlockCVE-2025-32424 · GHSA-7g34-7fvq-xxq62026-06-11See advisory
CriticalThere is a DoS vulnerability in FileStoreBlock with StepThroughItemsBlockCVE-2025-32422 · GHSA-9fr4-9jj9-mhh62026-06-11See advisory
CriticalThere is a DoS vulnerability in LoopVideoBlockCVE-2025-32392 · GHSA-267x-8jx3-gg6w2026-06-11See advisory
MediumSendEmailBlock allows SSRF via user-controlled SMTP server, bypassing IP blocklistCVE-2026-33234 · GHSA-4jwj-6mg5-wrwf2026-05-13See advisory
MediumAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-32425 · GHSA-vw3v-whvp-33v52026-05-130.6.32
MediumCredit system bypassed via direct block execution in POST /api/blocks/{block_id}/executeCVE-2026-45023 · GHSA-8pjg-mfqm-vrhr2026-05-11See advisory
HighUnsafe Redis cache deserialization (pickle.loads) would allow backend RCE in case of cache poisoningCVE-2026-33233 · GHSA-rfg2-37xq-w4m92026-05-11See advisory
HighUnauthenticated DoS via Disk Space ExhaustionCVE-2026-33232 · GHSA-374w-2pxq-c9jp2026-05-11See advisory
HighAuthenticated Session Hijacking via IDORCVE-2026-30950 · GHSA-q58p-v9r9-7gqj2026-05-11See advisory
CriticalAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2026-26020 · GHSA-4crw-9p35-9x542026-02-120.6.48
MediumReDoS (Regular Expression Denial of Service) at Code Extraction BlockGHSA-m2wr-7m3r-p52c2026-02-10See advisory
MediumAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2026-260062026-02-100.6.32
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-32393 · GHSA-5cqw-g779-9f9x2026-02-050.6.32
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2026-22038 · GHSA-rc89-6g7g-v5v72026-02-040.6.46
CriticalAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-62616 · GHSA-ggc4-4fmm-9hmc2026-02-040.6.34
CriticalAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-62615 · GHSA-r55v-q5pc-j57f2026-02-040.6.34
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2026-24780 · GHSA-r277-3xc5-c79v2026-01-290.6.44
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents.CVE-2025-53944 · GHSA-x77j-qg2x-fgg62025-07-30See advisory
LowAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-31494 · GHSA-958f-37vw-jx8f2025-04-150.6.1
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-31491 · GHSA-ggcm-93qg-gfhp2025-04-150.6.1
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-31490 · GHSA-wvjg-9879-3m7w2025-04-140.6.1
HighAutoGPT versions 0.3.4 and earlier are vulnerable to a Server-Side Template Injection (SSTI) that could lead to Remote Code Execution (RCE).CVE-2025-10402025-03-200.4.0
HighA Server-Side Request Forgery (SSRF) vulnerability was identified in the Requests utility of significant-gravitas/autogpt versions prior to v0.4.0.CVE-2025-04542025-03-200.4.0
CriticalA command injection vulnerability exists in the workflow-checker.yml workflow of significant-gravitas/autogpt.CVE-2024-81562025-03-200.5.1
HighAutoGPT is a platform that allows users to create, deploy, and manage continuous artificial intelligence agents that automate complex workflows.CVE-2025-22603 · GHSA-4c8v-hwxc-23562025-03-100.4.2
CriticalA vulnerability in significant-gravitas/autogpt version 0.5.1 allows an attacker to bypass the shell commands denylist settings.CVE-2024-60912024-09-11See advisory
CriticalAutoGPT, a component of significant-gravitas/autogpt, is vulnerable to an improper neutralization of special elements used in an OS command ('OS Command Injection') due to a flaw in its shell command validation function.CVE-2024-18812024-06-060.5.1
HighAn OS command injection vulnerability exists in the MacOS Text-To-Speech class MacOSTTS of the significant-gravitas/autogpt project, affecting versions up to v0.5.0.CVE-2024-18802024-06-060.5.1
HighA Cross-Site Request Forgery (CSRF) vulnerability in significant-gravitas/autogpt version v0.5.0 allows attackers to execute arbitrary commands on the AutoGPT server.CVE-2024-18792024-06-06See advisory
LowAuto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model.CVE-2023-37275 · GHSA-r7f7-qrrv-3fjh2023-07-130.4.3
HighAuto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model.CVE-2023-37274 · GHSA-5h38-mgp9-rj5f2023-07-130.4.3
HighAuto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model.CVE-2023-37273 · GHSA-x5gj-2chr-4ch62023-07-130.4.3

CVEs matched by the NVD's product names and advisories published in the product's own repository, checked 2026-11-18. Where the vendor runs the service, its fix reaches you without an update.

Sources: GitHub Advisory Database, reviewed and malware advisories (CC BY 4.0, https://github.com/advisories); advisories published by Cursor (https://github.com/cursor/cursor/security/advisories); advisories published by OpenClaw (https://github.com/openclaw/openclaw/security/advisories); advisories published by vLLM (https://github.com/vllm-project/vllm/security/advisories); advisories published by Open WebUI (https://github.com/open-webui/open-webui/security/advisories); advisories published by n8n (https://github.com/n8n-io/n8n/security/advisories); advisories published by Flowise (https://github.com/FlowiseAI/Flowise/security/advisories); advisories published by Langflow (https://github.com/langflow-ai/langflow/security/advisories); advisories published by the MCP Python SDK project (https://github.com/modelcontextprotocol/python-sdk/security/advisories); advisories published by MLflow (https://github.com/mlflow/mlflow/security/advisories); CISA Known Exploited Vulnerabilities catalog (CC0). Data as of 2026-11-22. ShieldMap describes how each problem works and how to fix it; it does not publish exploit code. Report a correction to [email protected].

This product uses the NVD API but is not endorsed or certified by the NVD. Product facts from github.com, checked 2026-11-18. Logos are their owners' marks, shown only to identify the products.