All news with #secret exposure tag

🦙 A critical vulnerability in Ollama (CVE-2026-7482) allows unauthenticated attackers to upload a crafted GGUF model file and trigger an out-of-bounds heap read in the model quantization pipeline. The flaw can leak process memory — including system prompts, conversation history, environment variables, API keys, and other secrets — to remote servers. Update to Ollama 0.17.1 and restrict network access.

🔒 A high-severity vulnerability in the AI-powered development tool Cursor allows installed extensions to read sensitive credentials stored locally, researchers at LayerX report. The issue stems from Cursor keeping API keys, session tokens and cached configuration in an unprotected SQLite database rather than using OS keychains or encryption, and it does not restrict extension access. LayerX assigned the flaw a CVSS score of 8.2 and demonstrated silent exfiltration without user prompts. Cursor acknowledged the notice but said trust boundaries are the user's responsibility; as of 28 April 2026 the vulnerability remains unresolved.

🔐 Researchers disclosed a major data exposure at Moltbook on January 31, 2026, revealing 35,000 emails and 1.5 million agent API tokens across 770,000 agents. Private messages contained plaintext third-party credentials, including OpenAI API keys, creating what the article calls a toxic combination — cross-app permissions that compound risk. The piece urges shifting review from single apps to the bridges between them and highlights procedural controls and dynamic SaaS security platforms like Reco to monitor runtime trust relationships and revoke risky tokens before exfiltration.

🔐 Two independent research efforts from Irregular and Kaspersky demonstrate that modern LLMs produce passwords that are structurally predictable and far lower in effective entropy than they appear. Models often repeat the same strings across sessions and conform to human-like patterns that fool standard strength meters. Autonomous coding agents are embedding these credentials into configuration files and repositories, and conventional secret scanners lack the means to detect them. Organizations should audit codebases, rotate suspect credentials, and require explicit use of cryptographically secure RNGs for all generated secrets.

🔒 TeamPCP's March 2026 compromise of LiteLLM packages on PyPI injected infostealer malware into versions 1.82.7 and 1.82.8 that ran during installs and updates. The malware harvested plaintext SSH keys, cloud credentials (AWS, Azure, GCP), Docker configs, IDE and agent memory files, and other local secrets, exploiting transitive dependencies. PyPI removed the packages within hours, but many downstream packages would have triggered execution. Use ggshield, pre-commit hooks, and filesystem scanning to detect and contain local secrets.

🔓Anthropic confirmed that internal source code for its coding assistant Claude Code was inadvertently published after a packaging error when version 2.1.88 was released to npm. The package included a source map exposing nearly 2,000 TypeScript files and over 512,000 lines of code; the release has since been removed. Anthropic says no customer data or credentials were exposed and is implementing measures to prevent recurrence.

🔓 An Anthropic employee accidentally published a source map in a public npm package, which allowed the proprietary source for Claude Code to be reconstructed. Anthropic says this was a release packaging error and that no sensitive customer data or credentials were exposed, and that it is rolling out measures to prevent recurrence. Security experts warn that source maps reveal original code, comments, internal constants and prompts, making vulnerabilities and secrets easier to find; the same mistake reportedly occurred previously.

🚨 Anthropic says it accidentally published the closed-source Claude Code source when an NPM release (v2.1.88) included a 60MB cli.js.map file that embedded original sources. The reconstructed tree contains roughly 1,900 files and 500,000 lines of code, and the leak has spread across GitHub and other platforms. Anthropic confirmed no customer data or credentials were exposed, called the incident a packaging error caused by human mistake, and is issuing DMCA takedowns while rolling out measures to prevent recurrence.

🔒 GitGuardian's State of Secrets Sprawl 2026 shows leaks accelerated in 2025, uncovering 29 million new hardcoded secrets — a 34% year-over-year increase and the largest single-year jump recorded. The report highlights three core trends: AI-driven credential exposures, unexpectedly widespread internal-repo and collaboration-tool leaks, and persistent remediation failures. It urges a shift from detection to continuous non-human identity governance, secrets vaulting, and automated rotation to reduce attacker access.

🛡️ Cybersecurity researchers uncovered five malicious Rust crates on crates.io that posed as time utilities while exfiltrating .env files to attacker infrastructure. The packages—chrono_anchor, dnp3times, time_calibrator, time_calibrators, and time-sync—were published in late February and early March 2026 and used a lookalike domain to collect secrets. Affected users should assume possible compromise: rotate keys, audit CI workflows, and limit outbound access from build systems.

🔑 South Korea's National Tax Service (NTS) accidentally included a photograph in a press release that exposed a handwritten cryptocurrency mnemonic seed phrase next to a seized Ledger device. Within hours the wallet holding roughly 4 million PRTG tokens (about US $4.8M) was emptied. The NTS removed the release and issued an apology; the incident underscores that publishing a wallet's seed phrase instantly nullifies any cold-storage security.

🔓 South Korea’s National Tax Service inadvertently exposed the mnemonic recovery phrase of a seized Ledger hardware wallet in a press release, enabling an attacker to drain approximately $4.8 million in crypto. The assets were confiscated during raids on 124 high-value tax evaders, but photos released by authorities showed a handwritten seed phrase that was not redacted. On-chain analysis shows the attacker deposited ETH for gas and moved 4 million Pre-Retogeum (PRTG) tokens to a new address in three transactions. The NTS removed the press release, and it is unclear whether a formal investigation has been launched.

⚠️ Truffle Security found nearly 3,000 Google Cloud API keys (prefix "AIza") embedded in client-side code that can now authenticate to Gemini endpoints when a project enables the Generative Language API. Attackers scraping sites can use exposed keys to access uploaded files, cached contents, and make LLM calls that charge victims' accounts. Google says it has implemented measures to detect and block leaked keys and advises rotating and restricting exposed keys.

🔒 Researchers at Truffle Security discovered that Google Cloud API keys, historically described as simple billing identifiers (prefix Aiza), began functioning as authentication tokens for embedded Gemini AI instances. A Common Crawl scan in November found 2,863 live, publicly exposed keys, including from major firms and Google itself, which could be used to retrieve uploaded files, cached context, or to consume API quota and incur charges. Google confirmed the issue after disclosure, restricted affected keys, and advises administrators to audit and rotate keys.

🔓 Google Cloud API keys that were once treated as non-sensitive can now authenticate to the Gemini generative AI assistant, creating a new attack path where keys embedded in client-side JavaScript expose private assistant data. TruffleSecurity discovered nearly 2,800 live, publicly accessible keys across sectors — including financial firms and a Google product — by scanning the November 2025 Common Crawl. Attackers who copy exposed keys can call Gemini endpoints to retrieve data or generate costly API usage; developers should audit projects for the Generative Language API, rotate exposed keys immediately, and use detection tools to prevent abuse.

⚠️ Check Point researchers disclosed critical vulnerabilities, CVE-2025-59536 and CVE-2026-21852, in Anthropic’s Claude Code that allow remote code execution and theft of Anthropic API keys via malicious repository-level configuration files. The flaws can be triggered simply by cloning and opening an untrusted project; built-in mechanisms such as Hooks, MCP integrations, and environment variables may be abused to bypass trust controls, execute hidden shell commands, and redirect authenticated API traffic before user consent. Stolen keys can expose shared workspaces, modify or delete resources, and generate unauthorized costs, underscoring a shift in the AI supply chain threat model.

🔍 Intruder scanned 5 million applications for secrets in built JavaScript bundles and found over 42,000 exposed tokens across 334 secret types. Many were active, high-risk credentials — including 688 repository tokens (GitHub/GitLab) and API keys for project management tools like Linear, some granting full access to private repos and CI/CD secrets. Traditional scanners, SAST, and DAST missed many of these because the secrets were introduced during build and lived only in bundled front-end code. The research highlights the urgent need for SPA spidering and explicit bundle scanning to prevent production leaks.

🔓 Security researchers at Wiz discovered a public Supabase API key in Moltbook’s client-side JavaScript that granted unauthenticated read/write access to the production database. The misconfiguration—absence of Row Level Security (RLS) policies—exposed around 1.5 million agent tokens, roughly 30,000 email addresses and thousands of private messages. With write privileges an attacker could impersonate any agent, inject malicious content or prompt-injection payloads, and deface the site. Moltbook’s developer has since remediated the issue after multiple rounds of fixes with Wiz.

⚠️ Chainlit, a widely used open-source framework for building conversational AI chatbots, contained high-severity vulnerabilities that can expose arbitrary files and permit server-side request forgery, enabling data theft and lateral movement within compromised environments. Zafran Security identified two primary issues: CVE-2026-22218 (arbitrary file read, CVSS 7.1) and CVE-2026-22219 (SSRF with SQLAlchemy, CVSS 8.3). Both were responsibly disclosed on November 23, 2025 and patched in Chainlit 2.9.4 on December 24, 2025. Administrators should upgrade, audit deployments for misuse, and rotate any potentially exposed credentials.

🔒 Chainlit, a widely used framework for building conversational AI applications, contained two server-side vulnerabilities (CVE-2026-22218 and CVE-2026-22219) that allow authenticated users to read arbitrary files and trigger SSRF in affected deployments. The flaws stem from insufficient validation of user-controlled properties in custom elements and SQLAlchemy-backed storage. Combined, they can expose environment variables, cached prompts, API keys and cloud metadata, enabling lateral movement beyond the app layer. Chainlit released 2.9.4 on 24 December 2025 and users are advised to apply the patch immediately; temporary WAF signatures were published as mitigation.