OpenAI
CompanySecurity timeline
- Security update reported by Trail of BitsTrail of Bits uses Codex to speed up backporting security fixes to stable Linux releases
- Phishing reported by Cyber Security NewsFake ChatGPT Plus billing emails steal card details and one-time passwords
- Vulnerability reported by GBHackersChained flaw in ChatGPT public share links exposed private conversations
- Vulnerability reported by Cyber Security NewsCritical ChatGPT flaw let one malicious link access private chats; OpenAI has fixed it
- Advisory or policy news by EFF DeeplinksEFF Warns Narrowing Fair Use for Generative AI Training Would Harm the Public
- Malware reported by Security AffairsYoba malware hides a crypto miner behind fake ChatGPT and Claude desktop apps
- Activity reported by Dark ReadingGrey market relay service disables OpenAI Codex and Claude Code sandboxes and reroutes enterprise data
- Vulnerability reported by Help Net SecurityEmbedded browsers in five AI coding assistants allow code execution, file theft and address spoofing
- Activity reported by GBHackersUnauthorised Codex and Claude Code relay service disables sandboxing at 12 organisations
- Advisory or policy news by Just SecurityAnalysis argues current liability law may not curb harms from rogue AI agents
- Malware reported by Cyber Security NewsNIK C2 malware abuses OpenAI API for credential theft and remote control
- Vulnerability reported by NCSC (UK)Critical PHP flaws CVE-2026-109618 and CVE-2026-109619 now exploited in the wild
- Security update reported by Chrome ReleasesGoogle releases Chrome 156 stable security update fixing 52 flaws, including critical bugs
Recent news · 15 stories
Trail of Bits uses Codex to speed up backporting security fixes to stable Linux releases
Trail of Bits describes using Codex agents to prepare backports of security fixes to stable Linux kernels, Ubuntu LTS and Fedora packages, with human engineers reviewing and submitting patches. By 29 October the project had helped deliver fixes for 149 distinct CVEs into stable kernel releases. The authors argue that thousands of known CVEs remain unfixed in stable branches, including a critical TCP stack flaw that had not reached distribution kernels weeks after upstream patching.
Fake ChatGPT Plus billing emails steal card details and one-time passwords
Fortra researchers documented a phishing campaign impersonating OpenAI billing notices that lead to fake checkout pages on short-lived WordPress sites. The pages capture card details, then hold victims on an indefinite spinner while a human operator, polling a dashboard, decides whether to request a one-time password, show an error or end the session. Fortra observed phishing only, with no malware delivery.
Chained flaw in ChatGPT public share links exposed private conversations
Researchers chained several flaws so that a public ChatGPT share link ran attacker-controlled JavaScript inside the victim's authenticated session, putting private conversations at risk. OpenAI fixed the issue on October 26 and classified it as P1, and the Parameter researchers' write-up says it earned a $5,000 bounty.
Vendor article argues penetration testing firms should run AI in-house to protect client data
The article, published as promotional content for PentestPad, says penetration testing firms face pressure to adopt AI and that staff often use it without approval. It argues that uploading client scope documents, vulnerabilities and credentials to cloud AI providers risks breaching confidentiality and compliance obligations, and recommends self-hosted models. PentestPad is presented as a reporting platform that can run on self-hosted or air-gapped infrastructure and exposes its data to agents through MCP.
Critical ChatGPT flaw let one malicious link access private chats; OpenAI has fixed it
A critical ChatGPT vulnerability allowed attacker-controlled JavaScript to run in a victim's signed-in session via a malicious shared conversation. It could list and read private conversations and exfiltrate data to an external server, with a single link sufficing in the demonstration. OpenAI fixed it on 26 October 2026 after a report from Parameter, which received a $5,000 bounty.
Byoky: open-source wallet keeps AI API keys encrypted on the user's device
Byoky is an open-source wallet for AI API keys from OpenAI, Anthropic, Google Gemini and other providers, available as Chrome and Firefox extensions and as iOS and Android apps. Keys are encrypted on the user's device, and the wallet sends provider requests on behalf of apps, which receive only a session token while the key stays in the wallet.
EFF Warns Narrowing Fair Use for Generative AI Training Would Harm the Public
The EFF has filed an amicus brief in In re OpenAI, arguing that narrowing the fair use doctrine would affect far more than generative AI developers. It warns that requiring lawful acquisition of training data or mandatory licensing would entrench large tech firms, impede research and journalism, and cast doubt on routine copying in everyday internet use.
Yoba malware hides a crypto miner behind fake ChatGPT and Claude desktop apps
Check Point Research tracks a campaign called Yoba that uses fake GitHub projects posing as OpenAI and Anthropic desktop apps, delivered via SEO poisoning. Each ZIP pairs a signed Setup.exe with a malicious winmm.dll for DLL sideloading, which installs a multi-stage loader that fetches payloads via Telegram, DNS-over-HTTPS and Catbox, and ends with an XMRig miner routed through Tor.
Grey market relay service disables OpenAI Codex and Claude Code sandboxes and reroutes enterprise data
Symantec's Threat Hunter Team identified a relay service active since at least early 2026 that has touched about a dozen organisations, mostly in mainland China. Users install the genuine Codex or Claude Code tool, then run a fileless setup script from a lookalike domain that rewrites the config to point at relay servers, stores the relay key as a user-level variable, sets sandbox mode to danger-full-access, enables network access and turns off telemetry. This is not a flaw in either product, but the relay can see every prompt, file and secret and could alter replies to run commands with the user's rights. No such abuse was observed.
Embedded browsers in five AI coding assistants allow code execution, file theft and address spoofing
Forever Security reported flaws in the embedded browsers of Cursor, GitHub Copilot, OpenAI Codex, VS Code and Claude Desktop, grouped under the name EMBRACE. Impacts include code execution in Cursor, address-bar spoofing in Copilot, and local file reads in Codex, VS Code and Claude Desktop, including SSH keys, .env files and npm tokens. Cursor and Copilot have been fixed, while Codex and VS Code were unpatched at publication.