OpenAI Reveals Six Model Incidents Involving Hidden Failures and Unauthorized Uploads

OpenAI on Wednesday disclosed six new instances of "unexpected or concerning model behavior" that took place over the past six months, while sharing a new framework for reporting, tracking, investigating, and disclosing model misalignment in a bid to improve transparency. "As AI systems grow more...

Inference-Engine Fingerprinting Attacks Are Practical: Exploring Model-Driven Environmental Discovery, Exploitation, and Escape

Frontier AI models are rapidly gaining the ability to exploit vulnerabilities in complex pieces of software. The risk is not theoretical, as evidenced by recent sandbox escapes performed by frontier models at OpenAI and Anthropic. Discussions of how to sandbox inference stack components often foc...

CVE-2026-61588

djust provides Phoenix LiveView-style reactive server-side rendering for Django with Rust-powered performance. Prior to version 1.0.7, when a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as...

CVE-2026-61588 djust's Django model serialization has no sensitive-field denylist: password hashes, privilege flags, and PII on a public view attribute are sent to the client

djust provides Phoenix LiveView-style reactive server-side rendering for Django with Rust-powered performance. Prior to version 1.0.7, when a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as...

CVE-2026-61588

djust , a Phoenix LiveView-style reactive server-side rendering framework for Django with Rust-powered performance, had a data-exposure flaw prior to version 1.0.7 . When a Django Model instance was assigned to a public view attribute , djust serialized the entire object to the client without a s...

EUVD-2026-81304

djust provides Phoenix LiveView-style reactive server-side rendering for Django with Rust-powered performance. Prior to version 1.0.7, when a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as...

CVE-2026-61588 djust's Django model serialization has no sensitive-field denylist: password hashes, privilege flags, and PII on a public view attribute are sent to the client

djust provides Phoenix LiveView-style reactive server-side rendering for Django with Rust-powered performance. Prior to version 1.0.7, when a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as...

GHSA-PVG3-6Q9J-MJ3X djust's Django model serialization has no sensitive-field denylist: password hashes, privilege flags, and PII on a public view attribute are sent to the client

Impact When a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as password the hash, privilege flags e.g. isstaff / issuperuser, tokens, and other PII to the browser. Because exposing model objec...

Exposure of Private Personal Information to an Unauthorized Actor

Overview djust is a Phoenix LiveView-style reactive components for Django with Rust-powered performance. Real-time UI updates over WebSocket, no JavaScript build step required. Affected versions of this package are vulnerable to Exposure of Private Personal Information to an Unauthorized Actor in...

djust's Django model serialization has no sensitive-field denylist: password hashes, privilege flags, and PII on a public view attribute are sent to the client

Impact When a Django Model instance is assigned to a public view attribute, djust serialized it to the client with no sensitive-field denylist — sending fields such as password the hash, privilege flags e.g. isstaff / issuperuser, tokens, and other PII to the browser. Because exposing model objec...

Out-of-bounds Write

Overview Affected versions of this package are vulnerable to Out-of-bounds Write in the parsing process of text models due to insufficient validation of child and split array values. An attacker can achieve arbitrary memory writes by supplying a specially crafted model file with invalid node...

CVE-2026-92786

LightGBM through 4.7.0 fails to validate child and split array values when parsing text models, allowing attackers to write out-of-bounds memory during SHAP prediction. Attackers can craft malicious model files with invalid node references that trigger out-of-bounds writes at attacker-chosen...

CVE-2026-92786 LightGBM through 4.7.0 Out-of-Bounds Write via Crafted Model

LightGBM through 4.7.0 fails to validate child and split array values when parsing text models, allowing attackers to write out-of-bounds memory during SHAP prediction. Attackers can craft malicious model files with invalid node references that trigger out-of-bounds writes at attacker-chosen...

CVE-2026-92786 LightGBM through 4.7.0 Out-of-Bounds Write via Crafted Model

LightGBM through 4.7.0 fails to validate child and split array values when parsing text models, allowing attackers to write out-of-bounds memory during SHAP prediction. Attackers can craft malicious model files with invalid node references that trigger out-of-bounds writes at attacker-chosen...

CVE-2026-92786 LightGBM through 4.7.0 Out-of-Bounds Write via Crafted Model

LightGBM through 4.7.0 fails to validate child and split array values when parsing text models, allowing attackers to write out-of-bounds memory during SHAP prediction. Attackers can craft malicious model files with invalid node references that trigger out-of-bounds writes at attacker-chosen...

CVE-2026-63127

A flaw was found in rmcp, a Rust SDK for the Model Context Protocol. The OAuth implementation in rmcp omits the RFC 9728 resource field from ResourceServerMetadata. This allows a malicious Model Context Protocol MCP server to publish metadata for a different legitimate MCP resource. Consequently,...

CVE-2026-61598 Client mass-assignment of arbitrary view attributes via the default dj-model update_model handler

djust provides Phoenix LiveView-style reactive server-side rendering for Django with Rust-powered performance. Prior to version 1.0.7, djust.mixins.modelbinding.ModelBindingMixin provides a default updatemodel event handler and is part of the LiveView base MRO, so every LiveView exposes it. It...

CVE-2026-92357

A vulnerability was identified in a2ui-project a2ui 0.8/0.9/1.0. Impacted is an unknown function of the file model-processor.ts of the component Model Processor. The manipulation of the argument currentsegment leads to information disclosure. The attack may be initiated remotely. The identifier o...

CVE-2026-92357 a2ui-project a2ui Model Processor model-processor.ts information disclosure

A vulnerability was identified in a2ui-project a2ui 0.8/0.9/1.0. Impacted is an unknown function of the file model-processor.ts of the component Model Processor. The manipulation of the argument currentsegment leads to information disclosure. The attack may be initiated remotely. The identifier o...

CVE-2026-92357

a2ui-project a2ui versions 0.8, 0.9, and 1.0 contain an information disclosure vulnerability in the Model Processor component (file model-processor.ts). The root cause is manipulation of the argument current[segment], which can be triggered remotely by a low-privileged user with no user interacti...