vLLM versions >= 0.10.2 and < 0.13.0 are missing sparse tensor validation in multimodal embeddings processing. Because PyTorch disables sparse tensor invariant checks by default, an attacker can submit crafted embedding requests with malformed (negative or out-of-bounds) tensor indices, when the prompt-embeds feature is enabled, to trigger crashes or resource exhaustion (denial of service), with potential for out-of-bounds/write-what-where memory corruption. This continues CVE-2025-62164, whose prior fix only disabled the feature by default rather than addressing the root cause.

A vulnerability was determined in assafelovic gpt-researcher up to 3.4.3. Affected is an unknown function of the component ws Endpoint. Executing a manipulation of the argument source_urls can lead to server-side request forgery. It is possible to launch the attack remotely. The exploit has been publicly disclosed and may be utilized. The project was informed of the problem early through an issue report but has not responded yet.

Capgo (Cap-go/capgo) before 12.128.2 contains an information disclosure vulnerability in the Supabase PostgREST SECURITY DEFINER RPC function public.rescind_invitation that allows unauthenticated attackers to enumerate organization existence. The function returns distinct error messages (NO_ORG vs NO_RIGHTS) when called with only a publishable API key, enabling attackers to discover valid organization IDs and increase the attack surface for targeted phishing or social engineering campaigns.

Capgo before 12.128.2 contains a denial of service vulnerability in the /auth/v1/otp endpoint that prevents email verification for two-factor authentication due to captcha validation failures. Authenticated users cannot complete 2FA enrollment as the backend consistently returns HTTP 500 errors with captcha verification process failed messages, blocking access to security controls.

Capgo before 12.128.2 contains an information disclosure vulnerability in the public.exist_app_v2 RPC function that allows unauthenticated attackers to enumerate app_ids by calling POST /rest/v1/rpc/exist_app_v2 with arbitrary appid parameters. Remote attackers can exploit this SECURITY DEFINER function to determine whether specific app_ids exist in the public.apps table, enabling cross-tenant app enumeration and privacy violations.

Capgo before 12.128.2 contains an information disclosure vulnerability in the unauthenticated /private/sso/check-domain endpoint that returns internal org_id and provider_id values. Attackers can enumerate email domains to build mappings of domains to organization UUIDs and SSO provider identifiers, enabling reconnaissance against Capgo tenants.

Capgo before 12.128.2 contains an authorization bypass vulnerability where write-scoped API keys can directly mutate protected channel configuration fields through PostgREST by exploiting a null authentication check in the immutability trigger. Attackers with write API keys can modify sensitive channel attributes such as public, allow_emulator, and security-related flags outside intended application routes.

Capgo before 12.128.2 lacks an UPDATE row-level security policy for the build_requests table, preventing API-key and anonymous access from persisting builder status updates. Attackers can exploit this missing policy to cause build status and error details to remain unpersisted, leaving build_requests rows stuck in pending state with null last_error values.

Capgo before 12.128.2 contains a server-side validation bypass vulnerability in organization security settings that allows authenticated org admins to persist invalid security policy state. Attackers can bypass backend validation by directly updating the public.orgs table from the browser, circumventing field-level validation checks for max_apikey_expiration_days and other security-sensitive configuration parameters.

Capgo before 12.128.2 contains an open redirect vulnerability in the confirm-signup endpoint that allows attackers to redirect users to arbitrary external websites. The confirmation_url parameter is not validated, enabling attackers to craft malicious links for phishing and credential harvesting attacks.

Capgo before 12.128.2 contains improper error handling in the /private/accept_invitation endpoint that returns HTTP 500 instead of safe 4xx errors when magic_invite_string is invalid. Attackers can trigger this vulnerability using only the public key by submitting malformed magic_invite_string values to cause server errors and leak internal processing details.

Capgo before 12.128.2 contains an open redirect vulnerability in stripe_portal and stripe_checkout endpoints that accept unvalidated callbackUrl, successUrl, and cancelUrl parameters. Authenticated attackers can craft malicious billing URLs to redirect users to attacker-controlled domains for phishing and credential harvesting.

A vulnerability was found in assafelovic gpt-researcher up to 3.4.3. This impacts an unknown function of the component HTTP REST API Endpoint. Performing a manipulation results in missing authentication. It is possible to initiate the attack remotely. The exploit has been made public and could be used. The project was informed of the problem early through an issue report but has not responded yet.

Capgo before 12.128.2 contains a cross-tenant preview namespace collision vulnerability caused by non-bijective decoding of double underscores to dots in preview hostname parsing. Attackers can register app IDs with underscores that collide with other tenants' dotted app IDs, causing preview misrouting and denial of preview access for victim applications.

Capgo before 12.128.2 allows multiple public channels for the same app and platform to coexist simultaneously, while unnamed /updates requests without defaultChannel implicitly resolve to a single hidden winner channel. An authorized app or channel manager can create ambiguous default update state and silently influence which bundle unnamed clients receive, breaking release routing integrity and predictability.

Capgo before 12.128.2 contains an information disclosure vulnerability in the public.invite_user_to_org RPC function that allows unauthenticated attackers to enumerate organization existence by observing distinct error responses. Attackers can call the SECURITY DEFINER function with a publishable API key to determine if an organization ID exists based on NO_ORG versus NO_RIGHTS responses, enabling tenant enumeration attacks.

Nuxt versions 4.0.0 before 4.4.7 and 3.x before 3.21.7 contain a server-side open redirect vulnerability in navigateTo that fails to properly validate path-normalized payloads like /..//evil.com and /.//evil.com. Attackers can bypass external-host checks using path-normalization techniques to redirect users to attacker-controlled sites via the Location header or meta-refresh, enabling phishing and OAuth authorization-code theft.

Capgo before 12.128.2 uses ILIKE pattern matching instead of exact matching for app_id lookup in the preview subdomain resolver, allowing underscore characters in app_id to act as SQL wildcards. Attackers can create apps with app_ids differing by one character at underscore positions to cause unintended pattern matches, breaking preview functionality for legitimate apps or causing app-id confusion.

Capgo before 12.128.2 contains a rate limit bypass vulnerability in the channel_self endpoint that allows attackers to circumvent rate limiting by rotating the user-controlled device_id parameter. Attackers can send multiple requests per second by changing device_id values to flood the channel_devices table and cause database exhaustion.

Capgo before 12.128.2 contains an information disclosure vulnerability in the /functions/v1/channel_self endpoint that allows unauthenticated attackers to enumerate non-public channel names and determine app existence and subscription status. Remote attackers can send GET requests with arbitrary app_id parameters to disclose internal rollout channels, enumerate valid applications across tenants, and leak billing status without authentication or device binding.