A security test showed that when a ChatGPT account had two connectors pointing to the same MCP server, the model confirmed an action's one-time code via one connector but redeemed it through the other, without informing the user. The server rejected the mismatched call because the code was cryptographically bound to the first connector's OAuth client, but the switch itself happened silently in every logged attempt.
community.openai.com
· 2026-11-06
Cisco Talos has detailed a self-deleting Windows infostealer dubbed VOLTA-AUTODOXER that harvests browser credentials, payment cards, email accounts, crypto wallets, Wi-Fi passwords and cloud configuration files, then compiles the data into a structured 'evidence bundle.' The malware sends this bundle to an AI model, Anthropic's Claude by default via OpenRouter, with a prompt to act as a forensic analyst and rank the victim's assets by cash-out value, saving the response as AI_Analysis.txt before exfiltrating everything and deleting itself.
tomshardware.com
· 2026-11-01
A new open-source macOS command-line utility called bigarrow lets AI agents overlay a big, click-through arrow, box or text label on screen to point at specific UI elements. Released under the MIT license as a Swift binary with an accompanying 'agent skill', it requires no special permissions, runs with no background daemon, and disappears automatically after use.
github.com
· 2026-10-09
Nudge Security reports that employees create an average of 88 OAuth app grants each, with 31 granting access to sensitive data. These grants often persist long after they are no longer needed, posing security risks. Managing and reviewing these permissions manually is increasingly difficult for IT teams.
bleepingcomputer.com
· 2026-10-08
NordLayer, citing its Browser Security Report 2026, says attackers increasingly compromise SaaS accounts via browser sessions—OAuth token theft, adversary-in-the-middle phishing, malicious extensions—without triggering endpoint detection and response (EDR) alerts because no malicious process or executable touches the host. It cites the 2025 Salesloft Drift breach, where group UNC6395 used stolen OAuth tokens to pull data from Salesforce via API calls, and a 2026 Microsoft-tracked campaign by Storm-2755 using search ads to phish Canadian employees. NordLayer's report found browser access present in all 504 reviewed applications, with 79% accessible only via browser.
bleepingcomputer.com
· 2026-10-02
F5 issued patches for a critical zero-day in its BIG-IP Access Policy Manager, tracked as CVE-2026-94127, which is being exploited to achieve remote code execution on systems configured as an OAuth Authorization Server. F5 says setups using APM only as an OAuth Client or Resource Server are unaffected, and offered an iRule-based mitigation for those unable to patch immediately. CISA has added the flaw to its Known Exploited Vulnerabilities catalog and directed federal agencies to remediate it by Friday.
bleepingcomputer.com
· 2026-09-23
French security firm CrowdSec disclosed that attackers used the Shai-Hulud worm to compromise a former employee's machine in May, stealing a GitHub OAuth token that still had read access to the company's private repositories. Over roughly nine minutes, attackers downloaded about 170 private repos plus 130+ public ones; CrowdSec only learned of the breach on September 16 after stolen source code surfaced on the cybercrime marketplace pwnforum.
darkreading.com
· 2026-09-22
BleepingComputer is partnering with Material Security for a live webinar on September 23 examining documented Google Workspace breaches. Speakers Rajan Kapoor of Material Security and Rick Fitzgerald of Fireside Consulting will dissect two incidents where attackers used social engineering and rogue OAuth apps to gain access, then walk through the response decisions that followed.
bleepingcomputer.com
· 2026-09-22
Security researchers warn that attackers can gain lasting access to SaaS and cloud accounts simply by tricking a logged-in user into approving a malicious OAuth application, sidestepping passwords, malware, and multifactor authentication entirely. Once granted, these app permissions can let attackers read email, browse files, pull source code, or touch CI/CD systems through legitimate API access until the tokens or grants are revoked.
darkreading.com
· 2026-09-18
BleepingComputer and Material Security are hosting a live webinar on September 23, 2026, examining real, publicly documented breaches of Google Workspace environments. Speakers Rajan Kapoor of Material Security and Rick Fitzgerald of Fireside Consulting LLC will dissect incidents involving social engineering and malicious OAuth apps to identify which defenses failed and which security controls actually matter.
bleepingcomputer.com
· 2026-09-18
Scry is a new search-over-the-web service built for AI agents, exposed through the Model Context Protocol as well as a direct HTTP API. Developers connect tools like Claude Code, Codex, or Cursor to a single MCP endpoint, authenticate via OAuth or an API key, and query a rolling crawl of high-information web pages using a SQL-like interface. The service also introduces congestion-based pricing tied to usage load.
scry.io
· 2026-09-17
BleepingComputer is hosting a live webinar on September 23, 2026, featuring Material Security's Rajan Kapoor and Fireside Consulting's Rick Fitzgerald, who will dissect real, publicly documented Google Workspace breaches. The session focuses on the critical early hours after a breach is discovered, including cases where attackers used social engineering paired with malicious OAuth apps to gain entry.
bleepingcomputer.com
· 2026-09-16
GoKawiil