CVE-2026-85695

FastChat (lm-sys/FastChat) has an authentication bypass in the /register_worker endpoint that allows unauthenticated attackers to register arbitrary worker addresses. This enables server-side request forgery (SSRF) and model spoofing : attackers can register malicious workers under victim model n...

CVE-2026-85694

LaVague 0.2.35 is vulnerable to remote code execution via its PythonFromMarkdownExtractor.extract_as_object function, which uses eval on untrusted language model output derived from web page content. An attacker can perform indirect prompt injection through a web page to inject malicious Python c...

CVE-2026-85692

Nightingale (n9e) contains an SSRF guard bypass in the isPublicIP function within aiagent/tools/http.go, the SSRF protection for the http_fetch AI-agent tool. The function only unwraps standard IPv4-mapped (::ffff:a.b.c.d) IPv6 addresses before checking them against a forbidden-range list, failin...

CVE-2026-85693

Chatbot UI (open-source chat application) contains an authorization bypass in its retrieval endpoint (/api/retrieval/retrieve). The root cause is a service-role Supabase client that bypasses row-level security and fails to validate file ownership, allowing any authenticated user to retrieve index...

CVE-2026-85691

MegaParse 0.0.55 contains an unauthenticated server-side request forgery (SSRF) in the POST /v1/url endpoint, which fetches caller-supplied URLs server-side without requiring authentication. Attackers can supply internal service URLs or cloud metadata endpoints and read their responses directly f...

CVE-2026-85690

Plandex 2.2.1 is vulnerable to a path traversal in the ApplyFiles function, which allows files to be written outside the project directory. Attackers can influence model output through poisoned repository files or attacker-controlled context to write to sensitive locations such as shell rc or cro...

CVE-2026-85689

llmware 0.4.6 contains an SQL injection vulnerability in its collection-database layer (llmware/resources.py). Filter and lookup values are directly string-interpolated into SQL WHERE clauses without parameterization or escaping, affecting both the SQLite and PostgreSQL backends. The filter valid...

CVE-2026-85688

TEN Framework 0.11.71 is vulnerable to unauthenticated arbitrary file read and write via the TMAN Designer file-content API endpoints (/api/designer/v1/file-content). Attackers can send POST and PUT requests to read arbitrary files or write malicious content to system paths, enabling code executi...

CVE-2026-85687

CVE-2026-85687 affects the surya 0.22.1 screenshot server, where the /info, /page, and /process routes accept raw file_path parameters without validation. The unsanitized paths are passed directly to Image.open or pypdfium2.PdfDocument , allowing an unauthenticated attacker to read any image or P...

CVE-2026-85685

AgentScope through 2.0.7.post1 is affected by a path traversal vulnerability in LocalWorkspace.add_skill . The skill_path request parameter is not confined, allowing an unauthenticated attacker over the network to supply an arbitrary directory path and copy files from any server location into the...

CVE-2026-85686

ms-swift 4.5.2 (ModelScope) contains an unauthenticated SSRF in the swift deploy OpenAI-compatible API. The vulnerable component is vision_utils.py, which fetches multimodal media URLs (image_url , audio_url , video_url ) without URL validation or redirect filtering. An attacker can supply arbitr...

CVE-2026-85684

marker through version 2.0.0 contains a path traversal vulnerability in its FastAPI /marker/upload handler. The root cause is a failure to sanitize the file.filename parameter, allowing unauthenticated attackers to supply filenames containing directory traversal sequences. This permits writing ar...

CVE-2026-85676

Dub , a link-shortening service, contains an open redirect vulnerability in its redir_url query parameter. The parameter is accepted on every short link without validation or domain-allowlist enforcement, allowing attackers to append an arbitrary external URL and redirect visitors through the tru...

CVE-2026-85675

OWL 's DocumentProcessingToolkit contains a server-side request forgery (SSRF) vulnerability in the extract_document_content tool. The tool fetches caller-supplied URLs without scheme, host, or IP filtering , allowing attackers to inject malicious URLs via prompt injection and cause the server to...

CVE-2026-85674

aider (aider-chat) is a CLI-based AI coding assistant that automatically loads a .aider.conf.yml configuration file from the root of the git repository it is launched in. A crafted repository can set test-cmd (executed at startup) or lint-cmd (executed on the first file edit), both of which are e...

CVE-2026-85673

LLaMA-Factory contains a server-side request forgery (SSRF) vulnerability in its OpenAI-compatible API multimodal media URL handler . The root cause is that the check_ssrf_url guard validates a URL only once, while requests.get subsequently follows HTTP redirects and re-resolves DNS without re-va...

CVE-2026-85671

QAnything 2.0.0 contains an authentication bypass in the /api/local_doc_qa/get_file_base64 and /api/local_doc_qa/get_doc endpoints. Unauthenticated attackers can enumerate file identifiers and retrieve base64-encoded files or parsed document chunks without ownership verification, enabling cross-t...

CVE-2026-85672

zerox 1.1.20 (getomni-ai) is affected by an OS command injection in its file download mechanism. The temporary file extension derived from document URLs is interpolated unsanitized into shell commands executed by poppler utilities. An attacker can craft a document URL with a malicious file extens...

CVE-2026-85670

CVE-2026-85670 affects Hugging Face tokenizers (observed in version 0.23.1 ). The vulnerability is an out-of-bounds buffer access in BpeBuilder::build (tokenizers/src/models/bpe/model.rs). When loading a tokenizer.json via Tokenizer::from_file/from_str, the builder sizes a scratch buffer to the l...

CVE-2026-85669

potpie through 2.0.0 fails to verify user ownership on the POST /conversations/{conversation_id}/code-changes/sync endpoint. Authenticated attackers can write arbitrary file changes into other users' conversations by supplying conversation IDs, resulting in unauthorized modification of pending co...