OpenAI’s agent escaped its sandbox during a security test

During an internal OpenAI security evaluation, a chain of AI models escaped its sandbox, reached the internet, and then accessed Hugging Face infrastructure to complete the test objective. OpenAI is a leading artificial intelligence AI research and deployment company. Its best-known product is...

ChatGPT AgentForger Flaw Could Deploy Rogue Workspace Agents via a Phishing Link

Cybersecurity researchers have disclosed a critical vulnerability in OpenAI's ChatGPT Workspace Agents that could have allowed a single phishing link to stealthily build, authorize, and deploy an autonomous artificial intelligence AI agent inside a victim's organization. The vulnerability has bee...

PT-2026-64318

Jan through 0.8.4, fixed in commit 3e1c1e7, contains a CORS misconfiguration vulnerability in its local API server that allows network-adjacent attackers to bypass trusted host restrictions by exploiting the server's replacement of user-configured trusted hosts with a wildcard that reflects...

CVE-2026-65700 h2oGPT 0.2.1 Path Traversal via OpenAI-compatible Files API

h2oGPT through 0.2.1 contains a path traversal vulnerability in the OpenAI-compatible files API that allows unauthenticated remote attackers to read, write, and delete arbitrary files accessible to the server process by supplying traversal sequences in the bearer token. The getuserdir function in...

n8n: Custom Header Credential Values Leaked in Plaintext into LLM Node Execution Data

Impact Custom HTTP headers configured in credentials for certain LLM sub-nodes including OpenAI, Anthropic, and Lemonade are masked in the n8n UI but are written in plaintext into execution data during workflow runs. Any authenticated user with access to the execution data for an affected workflo...

OpenAI Models Breached Hugging Face During Internal Cyber Test

OpenAI models escaped from a controlled cyber test, exploited zero-day flaws and breached Hugging Face while searching its production database for test answers...

OpenAI Says Its AI Models Escaped Sandbox, Targeted Hugging Face to Cheat Benchmark

OpenAI on Tuesday said a combination of its artificial intelligence AI models, including GPT-5.6 Sol and an "even more capable pre-release model," was behind the security incident that targeted Hugging Face's production infrastructure last week. The AI company said the models were operating with...

CVE-2026-63764

LMDeploy prior to 0.14.0 contains a server-side request forgery (SSRF) in the _load_http_url function of the media handler. The private-IP guard validates only the original URL; after HTTP redirects, hosts are not re-validated. An unauthenticated attacker can submit a crafted image_url to the cha...

PT-2026-62099

lmdeploy's OpenAI-compatible API server contains a server-side request forgery vulnerability that allows unauthenticated attackers to access internal services and cloud metadata endpoints by supplying a crafted image url that redirects to internal targets. Attackers can send a POST request to the...

MAL-2026-10732 Malicious code in hehehe (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 588dd776720f805d7d84a58c93ffcd21ed860e3fc21a48787d732eb6180b974d Package is published as an 'enterprise Windows Diagnostic Utility' but is an anti-proctoring cheating tool that also steals the installer's browser...

Malicious code in hehehe (npm)

--- -= Per source details. Do not edit below this line.=- Source: amazon-inspector 588dd776720f805d7d84a58c93ffcd21ed860e3fc21a48787d732eb6180b974d Package is published as an 'enterprise Windows Diagnostic Utility' but is an anti-proctoring cheating tool that also steals the installer's browser...

EUVD-2026-44953

text-generation-inference through 3.3.7 contains a server-side request forgery SSRF vulnerability in the OpenAI-compatible multimodal chat completions endpoint that allows unauthenticated network attackers to coerce the server into issuing arbitrary HTTP GET requests by supplying a crafted imageu...

[SECURITY] Fedora 43 Update: python-tiktoken-0.13.0-2.fc43

tiktoken is a fast BPE tokeniser for use with OpenAI's models...

[SECURITY] Fedora 44 Update: python-tiktoken-0.13.0-2.fc44

tiktoken is a fast BPE tokeniser for use with OpenAI's models...

CVE-2026-47475

NVIDIA TensorRT-LLM contains a vulnerability in the OpenAI-compatible inference API where an attacker could trigger a reachable assertion in the sampler thread. A successful exploit of this vulnerability might lead to denial of service...

CVE-2026-24271

NVIDIA TensorRT-LLM contains a vulnerability in the OpenAI-compatible inference API, where an attacker could cause allocation of GPU resources without limits or throttling. A successful exploit of this vulnerability might lead to denial of service...

CVE-2026-24271

NVIDIA TensorRT-LLM contains a vulnerability in the OpenAI-compatible inference API, where an attacker could cause allocation of GPU resources without limits or throttling. A successful exploit of this vulnerability might lead to denial of service...

CVE-2026-24271

NVIDIA TensorRT-LLM’s OpenAI-compatible inference API (CVE-2026-24271) is vulnerable to unthrottled GPU resource allocation, potentially causing denial of service. The affected product is NVIDIA TensorRT-LLM; the issue stems from uncontrolled resource allocation within the inference API. The secu...

CVE-2026-47475

NVIDIA TensorRT-LLM contains a vulnerability in the OpenAI-compatible inference API where an attacker could trigger a reachable assertion in the sampler thread. A successful exploit of this vulnerability might lead to denial of service...

CVE-2026-47475

NVIDIA TensorRT-LLM contains a vulnerability in the OpenAI-compatible inference API where an attacker could trigger a reachable assertion in the sampler thread. A successful exploit of this vulnerability might lead to denial of service...