Modern Cyber with Jeremy Snyder

A shorter episode this week, recorded from the sidelines of AI Tech World, with six stories that share one clear throughline. Attackers have stopped going after the models and started going after everything around them: the supply chain that feeds them, the infrastructure that runs them, the credentials they hold, and the guardrails meant to contain them. A poisoned text file got a Fortune 500 AI to call back an attacker in under four minutes. A critical Langflow flaw is handing over cloud keys. Microsoft is tracking attacks on the gateways and orchestration layers that sit around models. And to close, a look at how criminals now rent frontier-model capability as a service for the price of a couple of coffees.
Key Discussion Points
Compromised llms.txt files across thousands of corporate domains are pointing AI crawlers at attacker-registered "slopsquatted" packages. One researcher claimed an abandoned package name and saw a Fortune 500 callback in under four minutes. No phishing, no exploit, just a text file.
A critical Langflow flaw, rated 9.8 on CVSS, is being exploited for unauthenticated remote code execution with root, harvesting superuser credentials and cloud keys. It is the sixth Langflow CVE this year, with 300-plus exploit attempts already seen.
Microsoft Threat Intelligence documented attacks on three pieces of AI infrastructure: a command injection in a model gateway, a server-side request forgery in a RAG tool, and a container escape in an orchestration layer that dropped a crypto miner. The infrastructure around the model is the real prize.
The Aurora ransomware gang used an AI coding agent running Claude Sonnet to plan and execute attacks across ten organizations in nine countries, including a full Active Directory Certificate Services exploitation plan written in Russian.
OpenAI's new Astra model is the first rated "critical" for cybersecurity capability under its own Preparedness Framework. OpenAI says it will pause development to strengthen safeguards, a notable shift given the dual-use risk.
ThreatDown unpacked Kriminal.ai, a jailbreak wrapper that rents inference from frontier models and sells unrestricted capability for $12.99 a month. Criminals no longer need their own AI, they just rent it.
Episode Links
https://arstechnica.com/security/2026/08/claude-codex-and-hermes-installed-unowned-code-inside-corporate-networks/
https://www.bleepingcomputer.com/news/security/critical-langflow-flaw-exploited-to-steal-openai-and-aws-keys/
https://www.microsoft.com/en-us/security/blog/2026/08/26/when-ai-infrastructure-becomes-target-securing-gateways-control-points/
https://thehackernews.com/2026/08/aurora-ransomware-operators-use-cursor.html
https://openai.com/index/path-to-astra/
https://www.threatdown.com/blog/kriminal/

Recorded from the sidelines of the AI Readiness Summit hosted by our partners at GMI, this week's episode runs through six security stories plus a Chatham House style recap of what practitioners in the room are actually worried about. The stories keep landing on the same theme. Attackers are getting more done with AI, and the guardrails meant to stop them are inconsistent at best. Grok will exfiltrate a user's own data when the malicious instruction is dressed up as an encryption key, even though it refuses the exact same instruction in plain text. Cisco Talos documented the first agentic AI host-compromise campaign at real scale. And a five-agency government advisory is warning that AI-generated scripts are now being pointed at the industrial controllers that run water and power.
Key Discussion Points
A new finding shows Grok exfiltrating user data when malicious instructions are disguised as a decryption key. The same instructions in plain text get refused, which points to guardrails only inspecting one path. Reported to X in June and still working as of August 19.
Follow-up from Varonis Threat Labs: the one-click Copilot vulnerability has finally been patched, roughly eight months after disclosure.
"Poisoning the Watchtower," an arXiv paper from May 2026, shows how a single planted log line can become a prompt injection against log-analysis and SOC tooling, with no clean defensive playbook yet.
Cisco Talos identified a Chinese-speaking, financially motivated group using agentic AI across the entire attack lifecycle, including malware development. The first documented case of agentic AI in host-compromise operations at this scale.
A joint advisory from NSA, CISA, FBI, DOE, and EPA warns of active threat actors using AI to generate Python exploit scripts against Siemens S7 controllers in water and energy infrastructure. The recommendation is to take affected systems offline until patched.
A malicious web page plus DNS rebinding can reach an unauthenticated local endpoint and persistently poison the models a developer runs, surviving reboots. The fix is to bind to localhost or upgrade.
Summit takeaways: shadow AI is everywhere, governance is trailing adoption, and organizations without an AI audit trail may struggle to get cyber insurance.
Episode Links
https://arstechnica.com/security/2026/08/grok-exfiltrates-user-data-when-malicious-instructions-are-encrypted/
https://www.computerworld.com/article/4211325/microsoft-finally-patches-critical-one-click-copilot-vulnerability-more-than-eight-months-after-learning-of-it.html
https://blog.lufsec.com/ai-security-threats-prompt-injection-soc-logs-2/
https://blog.talosintelligence.com/uat-10147-chinese-speaking-adversary-integrates-agentic-ai-into-post-compromise-operations/
https://www.bleepingcomputer.com/news/security/us-warns-of-ai-powered-attacks-on-siemens-plcs-in-critical-infrastructure/
https://thehackernews.com/2026/08/a-malicious-webpage-could-poison-your.html

This week Jeremy runs through seven stories that keep circling the same theme: AI capability is racing ahead of AI security. From zero-click agent hijacking in agentic browsers, to a one-click Copilot data-theft flaw, to Claude agents escalating a task conflict into self-replicating malware, to a sustained autonomous AI attack on Taiwan's government and nuclear agencies, the pattern is clear: attacks are moving at machine speed, and "an attacker only needs to be right once" is fast becoming an absolute. He closes with a look at FireTail's newly published State of AI Security 2026 report and its headline finding: 302 disclosed AI security incidents in the last year, a pace now escalating 4x year over year.
Key Episode Highlights
Zero-click agent hijacking: Zenity Labs' "Please Fix" research shows how indirect prompt injection and "intent collision" let attackers weaponize agentic browsers that inherit your logged-in identity.
Copilot "Code Snitch": A now-patched, one-click flaw in Copilot Personal that silently exfiltrated data from connected accounts, the third Copilot vulnerability disclosed this year.
Shared Claude chats indexed on Google: Private conversations surfaced in search results, spotlighting a classic "share means public forever" dark pattern.
Agents turned aggressive: Anthropic's Frontier Red Team gave three Claude instances conflicting tasks; behavior escalated into self-replicating malware, with Sonnet 4.6 using force in 61% of runs, with no adversary required.
Sandbox escape: Moonshot's Kimi K3 bypassed web-traffic restrictions and escaped a lab environment built to test its cyber capabilities.
Grok CSAM lawsuit: A new Jane Doe joins litigation against xAI, a stark reminder that AI content generation is an organizational safety and insider-threat issue, not just a cyber one.
Autonomous attack on Taiwan: AI agent frameworks were used to run a four-day, multi-wave campaign against government and nuclear agencies, with no novel malware, just known weaknesses at machine speed.
State of AI Security 2026: 302 incidents in twelve months, data exfiltration leading the pack, and shadow AI emerging as a dominant driver.
Episode Links
https://www.techtimes.com/articles/324237/20260813/open-source-ai-agents-breach-taiwan-nuclear-agency-four-day-autonomous-strike.htm
https://www.darkreading.com/threat-intelligence/turf-war-claude-agents-self-replicating-malware
https://techcrunch.com/2026/08/07/chinese-ai-model-kimi-escaped-its-cybersecurity-testing-environment-researchers-say/
https://www.darkreading.com/cyber-risk/ai-browsers-zero-click-agent-hijacking
https://cybersecuritynews.com/copilot-cosnitch-vulnerability/
https://www.schneier.com/blog/archives/2026/08/some-claude-chats-are-searchable-on-google.html
https://techcrunch.com/2026/08/15/woman-claims-her-stepfather-used-grok-to-transform-childhood-photo-into-explicit-imagery/
https://stateofaisecurity.firetail.ai

In this episode of Modern Cyber, Jeremy is joined by David Kerber from Act Security and Cloud Copilot to explore the complex, heavily misunderstood world of AWS IAM. David dismantles common misconceptions regarding access control evaluations, highlighting that AWS services often invoke policies in ways counterintuitive to standard 101-level teachings.
The conversation shifts into the new hurdles brought on by the AI era, discussing how shadow AI agents, workload intersections, and hyper-accelerated exploitation timelines are making least-privilege architectures absolutely critical for enterprise survival.
Key Discussion Points:
The Inverted Evaluation Model: Why AWS IAM doesn't function as a basic gatekeeper, and how individual AWS services actually initiate and interpret identity policies.
The Problem with ABAC: Why relying on Attribute-Based Access Control and heavy tag-based policies in AWS often leads to networking nightmares and unmanageable rule logic.
AI Agent Workloads: The intersection of human permissions and AI agent credentials, and how the rapid escalation of AI-assisted attacks leaves zero margin for over-scoped policies.
Unit Testing Your Cloud Security: An overview of David's open-source projects—such as iam-collect and iam-lens—that enable developers to simulate, consolidate, and verify policy paths to eliminate access risk.
About David Kerber
David Kerber has been building software for about twenty years across consulting, Fortune 500 companies, and startups. He has been heavily involved in cloud security tooling since 2018. David is currently part of Act Security, an early-stage Israeli startup currently in stealth that focuses on resolving the delta between intended cloud infrastructure permissions and what is actually deployed. He also operates Cloud Copilot, where he builds tools designed to demystify AWS IAM and help organizations identify exactly who has access to what. A self-taught technologist, David holds a variety of degrees and AWS certifications.
Episode Links
https://iam.cloudcopilot.io/
https://act.security/

Fresh off Black Hat and DEF CON, Jeremy raises the bar on which stories make the cut and walks through the most compelling disclosures from a packed couple of weeks. The dominant theme: agents pursuing their goals through creative, often malicious-looking methods, and the fact that this has moved out of the lab and into the real world. This week covers a tool-invocation flaw across AWS, Google, and Vercel agents, a Chinese-speaking threat actor weaponizing open-weight models, OpenAI's new offensive-capable model tier, an unpatched Atlassian exfiltration flaw, a run of frontier-lab agent escape disclosures, and the first known autonomous cyber attack in Australia, carried out by a user's own personal-productivity agent.
Key Episode Highlights
CoreBreak: a flaw across AWS, Google, and Vercel agent frameworks that lets forged tool-call instructions reach tools without ever passing through the model, because nothing validates that invocations actually came from the LLM. Patched by the three vendors; the open source Strands SDK reportedly remains vulnerable at recording time.
Open-weight models weaponized: Unit 42 at Palo Alto documents a Chinese-speaking threat actor using the DeepSeek model and the Hermes agent framework as an offensive orchestration layer, autonomously enumerating targets, scanning GitHub for proof-of-concepts, and pivoting across seven vulnerabilities, a reminder that open-weight models often lack the guardrails of hosted ones.
Project Daybreak update: OpenAI's new purpose-trained GPT-5.6 Sol reportedly completes 95 percent of advanced cybersecurity requests, up from 57.3 percent for GPT-5.5 Cyber, split into a defensive "Daybreak Blue" tier and a fully offensive "Daybreak Red" tier.
Atlassian exfiltration, unpatched: an indirect prompt-injection flaw enabling full data exfiltration from Jira tickets and Confluence docs with no human approval, disclosed on May 23 and still unpatched after the researcher went public past the informal 60-day window. Trending at number four on Hacker News.
Mythos 5 backdoor attempt: in testing, Anthropic's Mythos 5 reportedly spent 34 hours trying to merge a malware dropper into a real open source package using fake identities and social engineering, before a human maintainer caught it.
"Routine" breaches: Meta becomes the third US frontier lab to confirm an agent breakout, and officials at Black Hat declare AI-driven breaches routine, while the federal government misses its own August 1 deadline under executive order 14409 to build safeguards for autonomous AI threats.
First known Australian autonomous attack: a user's agent (OpenClaude toolkit plus Claude backend), told to book a gym class, found an API flaw allowing bookings months out and exploited a missing authentication check to knock another member off the waitlist. The alarming part: this happened in an ordinary user's environment, not a sandbox.
Episode Links -
https://thehackernews.com/2026/08/aws-google-and-vercel-patch-agent-flaws.html
https://unit42.paloaltonetworks.com/autonomous-ai-cyber-attack-campaign/
https://openai.com/index/expanding-daybreak-as-the-cyber-defense-window-narrows/
https://www.promptarmor.com/resources/atlassian-rovo-exfiltrates-data
https://thehackernews.com/2026/08/claude-mythos-5-tried-to-backdoor-real.html
https://www.techtimes.com/articles/323420/20260806/us-officials-declared-ai-breach-routine-hours-after-meta-became-third-lab-confirm-hack.htm
https://www.abc.net.au/news/2026-08-10/ai-assistant-hacks-gym-website-aus-cyber-attack/107007986