The Compliance Brief for AI-built apps
Every Tuesday, the AI and app security stories from the past week, and what they mean for an app built quickly with AI tools.
Apps built with AI tools fail in familiar ways: missing authorization, secrets in the client, and LLM features that trust whatever comes back. The Compliance Brief tracks the attacks that exploit exactly that, one email a week.
Below are the stories from recent issues that bear on AI and app security, each with the short version and a link to the full take.
Free weekly email
Get the next issue on Tuesday
Join the list and the next issue arrives Tuesday morning. Or read a few below first.
Free. One email every Tuesday from Jacob Masse, and nothing else: signing up here does not add you to any other sequence. One click unsubscribes.
Latest on AI security, LLM attacks and app security flaws
- Chainalysis breached by what it calls offensive agentic techniques
Blockchain analytics firm Chainalysis disclosed that an external actor used what it described as offensive agentic techniques to reach internal systems and tools. - CISA and the Cyber Centre on AI token spend as a resilience problem
CISA, the Canadian Centre for Cyber Security and partners published joint guidance on managing AI token consumption. - Unauthenticated RCE in PHP, exploited already
Two critical PHP vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619 and nicknamed php2shell, allow an unauthenticated attacker to take full control of an affected server. - A Claude Code bug that quietly reads developer credentials
Researchers disclosed a flaw in the preview pane of Claude Code Desktop, dubbed SneakPeek, that allows an attacker to steal credentials from a developer machine without any click. - OpenCode breach is a lesson in holding other people's API keys
OpenCode said an intruder reached a legacy system on the night of October 16 and may have exposed third-party AI provider keys that users had saved in OpenCode Console. - A single prompt could have taken over an AWS agent fleet
Researchers disclosed a now-patched flaw in AWS Bedrock AgentCore, dubbed AgentCorruption. - AWS agent tooling can leak credentials
Researchers reported flaws in the Amazon Bedrock AgentCore SDK that could allow an attacker to run commands inside AI sandboxes and reach AWS credentials. - Your AI agents are logging in as humans and SOC 2 cannot tell
A vendor-authored piece argues that AI agents often operate through human credentials, so actions taken by an agent look identical to actions taken by the person whose credentials it borrowed.
Every issue on VibeCoded
- Issue 13: Chainalysis breached by what it calls offensive agentic techniques
- Issue 12: CISA and the Cyber Centre on AI token spend as a resilience problem
- Issue 11: A Claude Code bug that quietly reads developer credentials
- Issue 10: A single prompt could have taken over an AWS agent fleet
- Issue 9: AWS agent tooling can leak credentials
- Issue 8: Your AI agents are logging in as humans and SOC 2 cannot tell
- Issue 7: A regulator has now logged an AI agent as the attacker
- Issue 5: AI coding agents are pulling packages nobody registered
- Issue 4: Two arrests in the TeamPCP open-source supply chain spree
- Issue 2: The LiteLLM fallout is a CI credential problem, not an AI problem
- Issue 1: Hidden prompt injection is showing up in "Ask AI" buttons on marketing pages
Every issue in full, including the stories outside AI security, LLM attacks and app security flaws, is in the archive on traztech.ca. Issues with nothing on AI security, LLM attacks and app security flaws are listed there and not here.
Questions
How often does The Compliance Brief arrive?
Once a week, on Tuesday morning. Each issue covers the past week in five stories or so, with what happened and a short take on what it means for founders shipping AI-built and vibe coded apps.
What does it cost?
Nothing. It is written by Jacob Masse, Principal at TrazTech Inc., which operates VibeCoded. There is no paid tier.
Will signing up here send me anything else?
No. The form on this page adds you to The Compliance Brief and nothing else. Downloading a checklist elsewhere on the site is a separate signup, and it says what it sends before you give an address.
How do I stop it?
Every issue ends with a one-click unsubscribe link, and it is honoured immediately. Replying to any issue also reaches Jacob directly.
Free weekly email
Get it every Tuesday
One email a week on AI security, LLM attacks and app security flaws. Free, and one click to leave.
Free. One email every Tuesday from Jacob Masse, and nothing else: signing up here does not add you to any other sequence. One click unsubscribes.