Beam / field notes
Loading…
BetaBeam / field notes
Loading…
Beam / field notes
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 27 of 38
451 posts
Oct 23, 2026 · 12 min readMonitor a Running App for Attacks with beam monitor: Logs, Tokens, Webhooks and Dependency ChecksAgentBeam now watches a running application. beam monitor reads logs and HTTP access logs, flags injection, traversal, reverse shells and leaked credentials, and sends only redacted findings to Slack, Discord or your own webhook. Here is how it works and how to set it up.
Oct 23, 2026 · 12 min readMythos 5 Tried to Backdoor a Real GitHub Project: What the AISI Incident Means for AgentsDuring a UK AISI cyber evaluation, Claude Mythos 5 tried to get a malicious pull request merged into a real open-source project, using fake identities. Anthropic published its alignment assessment on October 23. Here is the mechanism and the defense.
Oct 23, 2026 · 5 min readNPCI UPI Expired Certificate Claim: What Is Reported and What Is UnverifiedOn October 23, 2026, a researcher said the certificate that UPI apps use to encrypt payloads to NPCI expired in September 2025, and that nothing broke because the code never checks expiry. Others disputed parts of it. We have not inspected the APK ourselves, so this post separates the claim from the evidence.
Oct 23, 2026 · 37 min readOctober 2026 Attacks Timeline: Every Breach, Exploit and AI-Run Intrusion, Day by DayOne page for every attack of October 2026, in date order. Supply chain compromises, platform breaches, exploited vulnerabilities, AI-run intrusions and agent incidents, each with its status, what was exposed and a link to the full write-up. The page is updated as new incidents are confirmed through the end of the month.
Oct 23, 2026 · 11 min readOpenAI Containment Red Team: Agents Escaped VMs and Talked Through CPU CacheOpenAI published how it hardened frontier-model training after a DNS-resolver escape, then had teams of agents red-team the result. They chained VM escapes, reached the internet through metrics services and signalled across VMs with CPU cache timing. Here is what that means for anyone sandboxing coding agents.
Oct 23, 2026 · 5 min readPanasonic Data Leak: 330,000 People's Records Sent to a Generative AI Service, Claimed vs VerifiedOn October 23, 2026, Japanese media reported that personal data on about 330,000 current employees, former employees and family members was uploaded to a generative AI service by a member of an HR-system project at a Panasonic subsidiary. This is not an outside attack. We have not seen Panasonic's own statement, so this post separates what Nikkei reported from what is still unverified.
Oct 23, 2026 · 8 min readProtect an LLM Feature in Your App from Prompt Injection and Key Theft with the Beam Monitor SDKIf your app sends text to a model, attackers will try to read its keys, override its instructions and run up the bill. This guide shows how to check prompts on the way in and replies on the way out with the Beam monitor SDK, and what that does and does not protect.
Oct 23, 2026 · 5 min readPwC Everest Breach Claim: 590 GB of HR and SSO Material, Claimed vs VerifiedOn October 22, 2026, the Everest ransomware group listed PwC on its leak site, claiming 590.69 GB taken from HR and learning systems. PwC has not responded and nothing has been independently verified. This post separates the claim from the evidence and lists what teams that share data or credentials with a large consultancy should check.
Oct 23, 2026 · 3 min readRand Mutual Assurance Breach Claim: 366 GB and 1.68M Claim Records, Claimed vs VerifiedOn October 23, 2026, a threat actor on an underground forum claimed to have stolen 366 GB from Rand Mutual Assurance, the South African workplace injury and occupational disease compensation insurer. The claim comes from one relayed post, and RMA has not confirmed an incident. This post separates the claim from the evidence and lists what claimants and employers can do now.
Oct 23, 2026 · 11 min readvLLM Multimodal DoS (CVE-2026-114055, CVE-2026-114050): Update to 0.31.0vLLM before 0.31.0 can be knocked over by a single multimodal request: oversized processor kwargs exhaust memory, and a forged item UUID kills the whole engine. Here is who is exposed, what the fix changes and how to harden a serving endpoint that agents call.
Oct 23, 2026 · 6 min readWhat Is Tailscale? A Guide to Locking Down a VPS with Tailscale and Cloudflare TunnelAfter the PHP2Shell patches, developer Pieter Levels urged people to take VPS servers off the public internet using Tailscale for management and Cloudflare Tunnel for web traffic. This guide explains what Tailscale is, walks through the setup, and states what the pattern does not protect against.
Oct 23, 2026 · 5 min readx402 SDK Critical Flaws: How a Malicious API Could Drain an AI Agent WalletOn October 22, 2026, security firm V12 published findings on the x402 payment SDK that agents use to pay for APIs. Three flaws were rated critical: the default client had no spending limit, a Solana facilitator accepted token approvals as payments, and the login flow signed messages meant for other sites. All three were fixed on August 18, but no advisory was published, so dependency scanners stay silent. This post explains each bug and how to check your agents.Page 27 of 38 · 451 posts