Advisories
Security vulnerabilities, privacy incidents, safety concerns, and policy updates affecting LLMs and AI agents.
Security vulnerabilities, privacy incidents, safety concerns, and policy updates affecting LLMs and AI agents.
9 items
CVE-2026-77177 affects Open GenAI Stack (aka ogx-ai) dated 2026-06-11, as used in the Meta AI backend for WhatsApp and other products. Prompt injection that carries Jinja2 template syntax can trigger server-side expression evaluation without sanitization, allowing code execution.
The GPTranslate WordPress plugin, all versions up to and including 2.34.6, exposes sensitive information through the enqueue_frontend_scripts function (CVE-2026-89278). Unauthenticated attackers can fetch any public page, apply the inverse transformation bundled in the plugin's public JavaScript asset, and recover the plaintext third-party AI provider API key (OpenAI, DeepL, xAI/Grok, Gemini, Claude, or Google Cloud Translation). The exposure affects the default configuration (gpt-3.5-turbo in client mode) and all supported non-DeepSeek providers, while deepseek-* models and gpt-* models in server-proxy mode suppress key emission.
NSA, CISA, and FBI issued a joint Cybersecurity Advisory warning that China-based AI companies, including DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, are running industrial-scale distillation campaigns against U.S. frontier models such as variants of Claude, GPT, Gemini, and Grok. The companies reportedly extracted billions of tokens across millions of requests since at least late 2024, routing traffic through APIs, cloud providers, third-party aggregators, and proxy "transfer stations" to bypass geographic restrictions and terms of use.
CVE-2026-4538 affects PyTorch 2.10.0, specifically an unknown function in the pt2 Loading Handler component. Manipulation of this component leads to deserialization, and the attack can only be performed from a local environment. A public exploit is available, and the project was informed early through a pull request but has not responded; NVD has not yet provided an assessment, while VulDB rates it CVSS 4.0 4.8 MEDIUM.
CVE-2025-55554 describes an integer overflow in the component torch.nan_to_num-.long() in pytorch v2.8.0. It is classified as CWE-190 (Integer Overflow or Wraparound). NVD had not yet provided an assessment at the time of the source text.
CVE-2025-3001 is a vulnerability classified as critical in PyTorch 2.6.0, affecting the function torch.lstm_cell. The manipulation leads to memory corruption, and an attacker must approach locally. The exploit has been publicly disclosed and may be used.
CVE-2024-48063 affects PyTorch versions up to and including 2.4.1. The RemoteModule in the distributed RPC framework contains a deserialization flaw, classified as CWE-502, that can lead to remote code execution. Several parties dispute the report, stating that this behavior is intended in PyTorch distributed computing.
CVE-2024-31583 affects PyTorch before v2.2.0. The source reports a use-after-free vulnerability (CWE-416) in torch/csrc/jit/mobile/interpreter.cpp. The NVD assessment is not yet provided, so no CVSS severity is assigned.
CVE-2024-31580 affects PyTorch before v2.2.0. The source reports a heap buffer overflow in the component /runtime/vararg_functions.cpp, which an attacker can trigger with a crafted input to cause a Denial of Service (DoS). The weakness is classified as CWE-122, Heap-based Buffer Overflow.
Fix: The source links a commit to the pytorch/pytorch GitHub repository (b5c3a17c2c207ebefcb85043f0cf94be9b2fef81) listed as a patch, but does not state a fixed version or workaround in the text. N/A -- no mitigation discussed in source.
Fix: The authoring agencies recommend U.S. AI companies implement comprehensive detection and mitigation, monitoring anomalous prompts, accounts, networks, and behaviors, including subscription-to-usage ratios, maximum usage from new accounts, and enterprise-scale throughput patterns. They also recommend deploying targeted response changes, subtly altering responses for suspected malicious distillation attempts to reduce the payoff to attackers. The source lists a third immediate action that is not shown in the excerpt.
CISA Cybersecurity Advisories