Beam / field notes

What we’re seeing in the field.

What we're seeing as we watch AI coding agents work — flagged actions, skill/MCP scan findings, and how to keep Claude Code, Codex, and the rest honest without slowing them down.Page 6 of 21

All authorsYash ThakkerBeam TeamNewest firstOldest firstTitle A–Z

AllAI AgentsSecurityThreat IntelligenceSupply ChainIncident ResponseAI Agent SecurityClaude CodeAgent SecurityPrompt InjectionCredential TheftMCPOpenAIAnthropicVulnerabilityClaimed vs VerifiedCodex

251 posts

Oct 29, 2026 · 10 min readContractor's Claude Code Agent on a DOD Network: Claimed vs VerifiedAnonymous sources told WIRED that a contractor connected a personal Claude Code account to a DOD system, and the agent sent credentials and a network map to a third-party site. Accenture and the DOD dispute or decline to confirm. Here is the claimed-versus-verified breakdown and the checks it suggests for your own setup.Oct 29, 2026 · 6 min readAU Small Finance Bank Breach Claim: ScarlettGroup, KYC and Aadhaar Data, Claimed vs VerifiedOn October 29, 2026, Hackmanac reported that a group calling itself ScarlettGroup claims to have breached AU Small Finance Bank and listed KYC, Aadhaar, PAN, credit and loan records on a leak page with a negotiation countdown. The bank has only asked people to contact it by direct message. This post separates the claim from the evidence.Oct 29, 2026 · 10 min read26,400 Live Cloud and AI Keys Found in Public Gitee Repos: What Coding-Agent Users Should CheckIndependent researcher Ben Zimmermann scanned 7,256,531 public Gitee repositories and found 26,400 credentials that still authenticated, including thousands of AI provider keys. This post covers the method, the numbers, the limits of the claim, and what it means if a coding agent commits code for you.Oct 29, 2026 · 8 min readLinux CVE-2026-98432: TCP Use-After-Free Patch Status and the "MPTCP Only" Claim CheckedCVE-2026-98432 went viral on October 29 as a 9.8 kernel bug that only matters if you run MPTCP. The CVE text describes the ordinary IPv4 TCP handshake path. Here is what is fixed, what is still pending, and what we could not confirm.Oct 29, 2026 · 9 min readLinux KSMBD RCE: Two CVSS 9.0 Kernel Flaws Posted Publicly Before a FixTrend Micro ZDI posted two in-kernel SMB server (KSMBD) flaws rated CVSS 9.0 to public Linux lists on October 20-21, 2026, before released kernels carried a fix. Public proofs of concept show a kernel memory leak and crashes, not code execution. This post covers the mechanism, who is actually exposed, how to check, and what to do.Oct 29, 2026 · 9 min readMicrosoft October 2026 Patch Tuesday: 1,411 Flaws, One Exploited Windows Shell Zero-Day, What to Install FirstMicrosoft released fixes for a record 1,411 vulnerabilities on October 13, 2026. One Windows Shell bug that bypasses Mark of the Web was already exploited, and CISA gave federal agencies until October 16. Here is the short list to install first.Oct 29, 2026 · 10 min readOVHcloud CLI Compromised: GitHub Release Binaries v0.11.0 to v0.15.0 Swapped for an InfostealerOn October 19, 2026, an attacker used a hijacked OVHcloud developer GitHub account to overwrite release binaries for ovhcloud-cli v0.11.0 through v0.15.0 with a secrets stealer. The window was about 94 minutes. This is a developer-tool compromise, not a backdoor in OVHcloud hosting. This post covers the mechanism, who is affected, how to check, and what to rotate.Oct 29, 2026 · 8 min readTeamUpdraft Breach: Stolen UpdraftCentral Keys Pushed Backdoor Plugins to WordPress SitesTeamUpdraft says an attacker got into teamupdraft.com on October 17-18, 2026 and reached a customer database that held UpdraftCentral connection keys. Patchstack saw those keys used to upload backdoor plugins to connected WordPress sites over about 18 hours. This is a key compromise, not a code flaw in UpdraftPlus. Here is the mechanism, the timeline and the cleanup.Oct 29, 2026 · 10 min readAgentic Attacks: Wiz's Seven Fingerprints From Flowise, Java and MCP IntrusionsWiz's new report documents three intrusions where an AI agent did the hands-on work: an exposed Flowise server, a Java gateway and an MCP-fronted GraphQL backend. Here are the seven fingerprints it lists, the public cases behind them, and what to check on your own machines.Oct 29, 2026 · 10 min readZeabur Second Breach and Shutdown: 3.2 Million Environment Variable Records Read, Rotate Every SecretZeabur, a deployment platform, announced on October 17, 2026 that it is shutting down, then published a preliminary report on October 19 about a second breach in two months. The attacker read the platform database, including 3,218,965 environment variable records, and logged in as root on 101 GCP dedicated servers. This post covers the timeline, who is affected, the rotation order, and what it means for API keys used by AI coding agents.Oct 28, 2026 · 11 min read123 Reg "AI-Driven Cyberattack": What the Registrar Claims vs. What Is VerifiedUK registrar 123 Reg told customers this week that an "AI-driven cyberattack" read their contact and product data, and suspects the group behind the ARTEX attacks on South Korean banks. Here is what is confirmed, what rests on the company's word alone, and what to check.Oct 28, 2026 · 10 min readAnthropic's AI-Built Exploit Chains: Chrome 156, Linux MPTCP and Windows QUIC Fixes to Apply NowOn October 27, 2026 Anthropic described five attack chains an unreleased research model found and exploited in Chrome, Safari, Android, Windows and Linux. This post covers what was disclosed, which patches to apply today, and what faster exploit development means for machines that run coding agents.

Page 6 of 21 · 251 posts