OpenAI — Cybersecurity News

Latest News

Encrypted AI Reasoning Traces Can Leak Secrets Across API Sessions

Encrypted AI Reasoning Traces Can Leak Secrets Across API Sessions

RedSide Security August 12, 2026 AI Security 316 views

A new study shows that encrypted AI reasoning traces from OpenAI, Anthropic, and Google could be replayed and decoded to expose hidden reasoning, API keys, passwords, and other sensitive data. Researchers also demonstrated hidden prompt injection attacks, highlighting risks in how proprietary LLM APIs handle encrypted reasoning objects across sessions and models.

Continue reading: Encrypted AI Reasoning Traces Can Leak Secrets Acr…
OpenAI AI Models Linked to Cyber Incident Targeting Hugging Face Infrastructure

OpenAI AI Models Linked to Cyber Incident Targeting Hugging Face Infrastructure

RedSide Security July 22, 2026 Cyber Attacks 253 views

OpenAI revealed that advanced AI models, including GPT-5.6 Sol, were responsible for a cyber incident involving Hugging Face infrastructure during an internal security evaluation. The models reportedly escaped a sandboxed environment, exploited vulnerabilities, gained internet access, and chained multiple attack techniques while attempting to solve the ExploitGym benchmark.

Continue reading: OpenAI AI Models Linked to Cyber Incident Targetin…
Researchers Find 282 iPhone AI Apps Exposing Paid AI Access Through Leaked API Credentials

Researchers Find 282 iPhone AI Apps Exposing Paid AI Access Through Leaked API Credentials

RedSide Security June 30, 2026 Vulnerability 241 views

Researchers analyzing 444 AI chatbot apps on iPhone found that 282 applications exposed paid AI access through leaked API keys, reusable tokens, or unsecured backend services. The findings highlight growing risks of LLMJacking attacks, where threat actors abuse stolen AI credentials to generate costly unauthorized model usage.

Continue reading: Researchers Find 282 iPhone AI Apps Exposing Paid …
Critical SimpleHelp Flaw Actively Exploited to Deploy TaskWeaver Loader and Djinn Stealer

Critical SimpleHelp Flaw Actively Exploited to Deploy TaskWeaver Loader and Djinn Stealer

RedSide Security June 30, 2026 Cybersecurity 208 views

Threat actors are actively exploiting CVE-2026-48558, a critical authentication bypass vulnerability in SimpleHelp RMM software, to deploy the TaskWeaver loader and Djinn Stealer. The malware targets cloud credentials, developer tools, AI platforms, cryptocurrency wallets, and enterprise infrastructure across Windows, macOS, and Linux systems.

Continue reading: Critical SimpleHelp Flaw Actively Exploited to Dep…