{"id":"CVE-2026-54249","aliases":["EUVD-2026-50562","GHSA-h7p7-w5gc-xj3w","PYSEC-2026-3692","PYSEC-2026-3693"],"status":"published","title":"VercelAIAdapter trusts client-controlled `providerMetadata` to construct `UploadedFile` — S3/GCS confused deputy via provider metadata injection","description":"Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.","severity":{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium","assessor":"cna","from":"cve"},"epss":{"day":"2026-11-10","score":0.00315,"percentile":0.24639,"model":"v2026.06.15"},"facts":{"in_kev":false,"known_exploited":false,"public_exploit":false,"fix_available":true},"cwes":["CWE-918"],"published":"2026-07-29T20:28:07Z","modified":"2026-10-28T02:30:46Z","fixes":[{"ecosystem":"PyPI","package":"pydantic-ai-slim","fixed_in":["1.106.0","2.0.0b6"],"sources":["ghsa","osv"]},{"ecosystem":"PyPI","package":"pydantic-ai","fixed_in":["1.106.0","2.0.0b6"],"sources":["ghsa","osv"]}],"affected":[{"source":"ghsa","ecosystem":"PyPI","package":"pydantic-ai-slim","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"pydantic-ai-slim","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"pydantic-ai","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"}]}]},{"source":"ghsa","ecosystem":"PyPI","package":"pydantic-ai","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai-slim","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"}]}],"versions":["1.100.0","1.101.0","1.102.0","1.103.0","1.104.0","1.105.0","1.65.0","1.66.0","1.67.0","1.68.0","1.69.0","1.70.0","1.71.0","1.72.0","1.73.0","1.74.0","1.75.0","1.76.0","1.77.0","1.78.0","1.79.0","1.80.0","1.81.0","1.82.0","1.83.0","1.84.0","1.84.1","1.85.0","1.85.1","1.86.0","1.86.1","1.87.0","1.88.0","1.89.0","1.89.1","1.90.0","1.91.0","1.92.0","1.93.0","1.94.0","1.95.0","1.95.1","1.96.0","1.96.1","1.97.0","1.98.0","1.99.0"]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai-slim","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}],"versions":["2.0.0b1","2.0.0b2","2.0.0b3","2.0.0b4","2.0.0b5"]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"}]}],"versions":["1.100.0","1.101.0","1.102.0","1.103.0","1.104.0","1.105.0","1.65.0","1.66.0","1.67.0","1.68.0","1.69.0","1.70.0","1.71.0","1.72.0","1.73.0","1.74.0","1.75.0","1.76.0","1.77.0","1.78.0","1.79.0","1.80.0","1.81.0","1.82.0","1.83.0","1.84.0","1.84.1","1.85.0","1.85.1","1.86.0","1.86.1","1.87.0","1.88.0","1.89.0","1.89.1","1.90.0","1.91.0","1.92.0","1.93.0","1.94.0","1.95.0","1.95.1","1.96.0","1.96.1","1.97.0","1.98.0","1.99.0"]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}],"versions":["2.0.0b1","2.0.0b2","2.0.0b3","2.0.0b4","2.0.0b5"]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"},{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}],"versions":["1.100.0","1.101.0","1.102.0","1.103.0","1.104.0","1.105.0","1.65.0","1.66.0","1.67.0","1.68.0","1.69.0","1.70.0","1.71.0","1.72.0","1.73.0","1.74.0","1.75.0","1.76.0","1.77.0","1.78.0","1.79.0","1.80.0","1.81.0","1.82.0","1.83.0","1.84.0","1.84.1","1.85.0","1.85.1","1.86.0","1.86.1","1.87.0","1.88.0","1.89.0","1.89.1","1.90.0","1.91.0","1.92.0","1.93.0","1.94.0","1.95.0","1.95.1","1.96.0","1.96.1","1.97.0","1.98.0","1.99.0","2.0.0b1","2.0.0b2","2.0.0b3","2.0.0b4","2.0.0b5"]},{"source":"osv","ecosystem":"PyPI","package":"pydantic-ai-slim","ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"1.65.0"},{"fixed":"1.106.0"},{"introduced":"2.0.0b1"},{"fixed":"2.0.0b6"}]}],"versions":["1.100.0","1.101.0","1.102.0","1.103.0","1.104.0","1.105.0","1.65.0","1.66.0","1.67.0","1.68.0","1.69.0","1.70.0","1.71.0","1.72.0","1.73.0","1.74.0","1.75.0","1.76.0","1.77.0","1.78.0","1.79.0","1.80.0","1.81.0","1.82.0","1.83.0","1.84.0","1.84.1","1.85.0","1.85.1","1.86.0","1.86.1","1.87.0","1.88.0","1.89.0","1.89.1","1.90.0","1.91.0","1.92.0","1.93.0","1.94.0","1.95.0","1.95.1","1.96.0","1.96.1","1.97.0","1.98.0","1.99.0","2.0.0b1","2.0.0b2","2.0.0b3","2.0.0b4","2.0.0b5"]}],"products":[{"vendor":"pydantic","product":"pydantic-ai"},{"vendor":"pydantic","product":"pydantic-ai-slim"},{"vendor":"pydantic","product":"pydantic_ai"}],"references":[{"url":"https://github.com/pydantic/pydantic-ai/security/advisories/GHSA-h7p7-w5gc-xj3w","tags":["x_refsource_CONFIRM","Vendor Advisory","advisory","web"]},{"url":"https://nvd.nist.gov/vuln/detail/CVE-2026-54249","tags":["advisory"]},{"url":"https://github.com/pydantic/pydantic-ai/releases/tag/v1.106.0","tags":["web"]},{"url":"https://github.com/pydantic/pydantic-ai/releases/tag/v2.0.0b6","tags":["web"]},{"url":"https://github.com/advisories/GHSA-h7p7-w5gc-xj3w","tags":["advisory"]},{"url":"https://github.com/CVEProject/cvelistV5/tree/main/cves/2026/54xxx/CVE-2026-54249.json","tags":["advisory"]},{"url":"https://github.com/pydantic/pydantic-ai","tags":["package"]},{"url":"https://pypi.org/project/pydantic-ai","tags":["package"]},{"url":"https://pypi.org/project/pydantic-ai-slim","tags":["package"]}],"sources":{"cve":[{"id":"CVE-2026-54249","title":"VercelAIAdapter trusts client-controlled `providerMetadata` to construct `UploadedFile` — S3/GCS confused deputy via provider metadata injection","description":"Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.","status":"published","published":"2026-07-29T20:28:07.154Z","modified":"2026-07-30T13:08:12.954Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium","assessor":"cna"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T13:37:58.613221Z"}],"euvd":[{"id":"EUVD-2026-50562","description":"Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.","status":"published","published":"2026-07-29T20:28:07Z","modified":"2026-07-30T13:08:12Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"updated_at":"2026-11-02T18:02:28.93005Z"}],"ghsa":[{"id":"GHSA-h7p7-w5gc-xj3w","title":"Pydantic AI: Unvalidated UploadedFile references in UI adapters allow server-side file access using the application's credentials","description":"### Summary\n\nA client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.\n\n### Details\n\nUI adapters reconstruct file parts from client-submitted message history and forward them to the model provider. File **URL** parts are validated against a scheme allowlist before being forwarded, but `UploadedFile` references — which point to a file by provider file ID or cloud-storage URI (e.g. `s3://…`, `gs://…`) — were forwarded without validation.\n\nBecause the provider resolves an `UploadedFile` using the server-side identity (IAM role, service account, or provider API key) rather than the client's, a client that crafts message history containing an attacker-chosen `UploadedFile` can cause the server to read objects belonging to its own account or to other tenants, given a referenceable identifier.\n\n### Impact\n\nApplications that pass untrusted client-submitted message history to an agent through a UI adapter (such as the Vercel AI adapter). Exploitation requires the attacker to reference a valid file identifier; depending on how the application names objects, such identifiers are not always unguessable.\n\n### Patches\n\nUpgrade to `1.106.0` (1.x) or `2.0.0b6` (the 2.x beta line), which validate `UploadedFile` references on client-submitted messages the same way file URLs are validated.\n\n### Workarounds\n\nIf users cannot upgrade, do not pass untrusted client-submitted message history to the agent, or strip `UploadedFile` parts from incoming messages before running the agent.","status":"published","published":"2026-08-13T14:16:26Z","modified":"2026-10-22T08:27:23Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T14:41:34.690842Z"}],"nvd":[{"id":"CVE-2026-54249","description":"Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.","status":"published","published":"2026-07-29T21:17:47.323Z","modified":"2026-08-04T13:22:03.63Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium","assessor":"secondary"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T14:12:33.629616Z"}],"osv":[{"id":"CVE-2026-54249","title":"VercelAIAdapter trusts client-controlled `providerMetadata` to construct `UploadedFile` — S3/GCS confused deputy via provider metadata injection","description":"Pydantic AI is a Python agent framework for building Generative AI applications. In versions 1.65.0 through 1.105.0, and 2.0.0b1 through 2.0.0b5, a client that submits message history to a Pydantic AI UI adapter (such as the Vercel AI adapter) can reference arbitrary files in the application's model-provider or cloud-storage account. While file URL parts are validated against a scheme allowlist, UploadedFile references — which point to a file by provider file ID or cloud-storage URI (e.g. s3://…, gs://…) — were forwarded without validation. Because the provider resolves an UploadedFile using the server-side identity (IAM role, service account, or provider API key) rather than the client's, an attacker can craft message history to make the server read objects from its own account or other tenants, given a referenceable identifier. Exploitation requires a valid file identifier, which is not always unguessable depending on how the application names objects. This issue has been fixed in versions 1.106.0 and 2.0.0b6.","status":"published","published":"2026-07-29T20:28:07.154Z","modified":"2026-10-28T02:30:46.433325504Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T15:45:10.818486Z"},{"id":"GHSA-h7p7-w5gc-xj3w","title":"Pydantic AI: Unvalidated UploadedFile references in UI adapters allow server-side file access using the application's credentials","description":"### Summary\n\nA client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.\n\n### Details\n\nUI adapters reconstruct file parts from client-submitted message history and forward them to the model provider. File **URL** parts are validated against a scheme allowlist before being forwarded, but `UploadedFile` references — which point to a file by provider file ID or cloud-storage URI (e.g. `s3://…`, `gs://…`) — were forwarded without validation.\n\nBecause the provider resolves an `UploadedFile` using the server-side identity (IAM role, service account, or provider API key) rather than the client's, a client that crafts message history containing an attacker-chosen `UploadedFile` can cause the server to read objects belonging to its own account or to other tenants, given a referenceable identifier.\n\n### Impact\n\nApplications that pass untrusted client-submitted message history to an agent through a UI adapter (such as the Vercel AI adapter). Exploitation requires the attacker to reference a valid file identifier; depending on how the application names objects, such identifiers are not always unguessable.\n\n### Patches\n\nUpgrade to `1.106.0` (1.x) or `2.0.0b6` (the 2.x beta line), which validate `UploadedFile` references on client-submitted messages the same way file URLs are validated.\n\n### Workarounds\n\nIf users cannot upgrade, do not pass untrusted client-submitted message history to the agent, or strip `UploadedFile` parts from incoming messages before running the agent.","status":"published","published":"2026-08-13T14:16:26Z","modified":"2026-08-19T12:55:44.916270032Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"cwes":["CWE-918"],"updated_at":"2026-11-02T15:46:48.920784Z"},{"id":"PYSEC-2026-3692","title":"Pydantic AI: Unvalidated UploadedFile references in UI adapters allow server-side file access using the application's credentials","description":"### Summary\n\nA client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.\n\n### Details\n\nUI adapters reconstruct file parts from client-submitted message history and forward them to the model provider. File **URL** parts are validated against a scheme allowlist before being forwarded, but `UploadedFile` references — which point to a file by provider file ID or cloud-storage URI (e.g. `s3://…`, `gs://…`) — were forwarded without validation.\n\nBecause the provider resolves an `UploadedFile` using the server-side identity (IAM role, service account, or provider API key) rather than the client's, a client that crafts message history containing an attacker-chosen `UploadedFile` can cause the server to read objects belonging to its own account or to other tenants, given a referenceable identifier.\n\n### Impact\n\nApplications that pass untrusted client-submitted message history to an agent through a UI adapter (such as the Vercel AI adapter). Exploitation requires the attacker to reference a valid file identifier; depending on how the application names objects, such identifiers are not always unguessable.\n\n### Patches\n\nUpgrade to `1.106.0` (1.x) or `2.0.0b6` (the 2.x beta line), which validate `UploadedFile` references on client-submitted messages the same way file URLs are validated.\n\n### Workarounds\n\nIf users cannot upgrade, do not pass untrusted client-submitted message history to the agent, or strip `UploadedFile` parts from incoming messages before running the agent.","status":"published","published":"2026-08-19T11:56:26.37585Z","modified":"2026-08-19T12:55:44.916270032Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"updated_at":"2026-11-02T15:52:56.847083Z"},{"id":"PYSEC-2026-3693","title":"Pydantic AI: Unvalidated UploadedFile references in UI adapters allow server-side file access using the application's credentials","description":"### Summary\n\nA client that can submit message history to a Pydantic AI UI adapter can reference arbitrary files in the application's model-provider or cloud-storage account. The server forwards the reference to the model provider, which fetches it using the server's own credentials, allowing the client to read files it should not have access to.\n\n### Details\n\nUI adapters reconstruct file parts from client-submitted message history and forward them to the model provider. File **URL** parts are validated against a scheme allowlist before being forwarded, but `UploadedFile` references — which point to a file by provider file ID or cloud-storage URI (e.g. `s3://…`, `gs://…`) — were forwarded without validation.\n\nBecause the provider resolves an `UploadedFile` using the server-side identity (IAM role, service account, or provider API key) rather than the client's, a client that crafts message history containing an attacker-chosen `UploadedFile` can cause the server to read objects belonging to its own account or to other tenants, given a referenceable identifier.\n\n### Impact\n\nApplications that pass untrusted client-submitted message history to an agent through a UI adapter (such as the Vercel AI adapter). Exploitation requires the attacker to reference a valid file identifier; depending on how the application names objects, such identifiers are not always unguessable.\n\n### Patches\n\nUpgrade to `1.106.0` (1.x) or `2.0.0b6` (the 2.x beta line), which validate `UploadedFile` references on client-submitted messages the same way file URLs are validated.\n\n### Workarounds\n\nIf users cannot upgrade, do not pass untrusted client-submitted message history to the agent, or strip `UploadedFile` parts from incoming messages before running the agent.","status":"published","published":"2026-08-19T11:56:26.304772Z","modified":"2026-08-19T12:55:44.916270032Z","cvss":[{"version":"3.1","score":6.8,"vector":"CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:C/C:H/I:N/A:N","level":"medium"}],"updated_at":"2026-11-02T15:52:56.847083Z"}]}}