SetupBots
Welcome to SetupBots

Sign in to download the installer, manage access, and keep your setup resources in one place.

Sign inCreate an account
  • Installer dashboard
  • Account settings
Start here
  • Getting Started with AIFREE
  • Advanced AI InstallationFREE
  • AI for Business OwnersFREE
  • Introduction to APIsFREE
  • Claude Connections: CLI, MCP & APIsFREE
  • AI Cold Email + Affiliate MasteryFREE
All starter courses
Featured courses

AI Websites Fundamentals

Free

AI Design Fundamentals

Free

Supabase from Zero

Free

AI SEO — Scale Traditional SEO with AI

Free

AI Search — Get Cited by ChatGPT, Claude, Perplexity

Free

Vercel Deploy & Production

Free
Browse the full catalog — free
SUPER/SETUP
The agent brain API.
  • Free during early access — all of it
  • Your agent queries verified setup methodology
  • Claude Code, Codex, Cursor, VS Code, Antigravity
Free with your account
No card · generous daily allowance
HomeCoursesAI Security & Privacy 101

AI Security & Privacy 101

NEWIntermediate2h 35mUpdated 2026-05-01
AboutWhat you'll learnCurriculumArtifactsInstructorFAQ

About this course

You're already coding with AI. You're already using it at work. This course is the difference between being a smart user and being the breach in next quarter's headlines. 6 modules · 35 lessons · real 2026 incidents · downloadable vendor-diligence checklist + contract clause library + 10 audit prompts you can run on your own codebase tomorrow.

What you'll learn

Spot prompt-injection attack patterns (direct + indirect) and harden against them in code
Build the two-layer rate-limit defense — Vercel Firewall + Supabase credits — that survives serverless cold starts
Run Supabase Security & Performance Advisors before every deploy (the built-in audit most devs miss)
Pick the right HIPAA route — AWS Bedrock + BAA, Anthropic Enterprise, or local (Google MedGemma, Amazon HealthLake AI)
Read SOC 2 / GDPR / WCAG 2.1 AA well enough to demand the right contract clauses
Audit licenses with `npx license-checker --failOn GPL-3.0;AGPL-3.0;SSPL-1.0` before every client delivery
Pen-test your own app WITH AI — Caido, PromptFoo, garak

Course curriculum

6 modules·35 lessons·Total 2h 35m
01The Mexican gov breach — one attacker, Claude Code + GPT-4.1, 195M taxpayer + 220M civil records (Dec 2025–Feb 2026)Preview2m
02The $12M bank loss — procurement shadow agent + indirect prompt injection auto-approved fraud (Q1 2026)Preview2m
03600+ firewalls in 55 countries — fully autonomous AI agent, no human operator2m
04Microsoft Semantic Kernel RCE — CVE-2026-25592 + CVE-2026-26030, prompt injection → host-level code execution (May 2026)2m
05The number: 88% of AI-deploying orgs had a confirmed or suspected security incident this year. Only 6% of security budgets fund AI.2m
01Myths — 'AI is smarter so it's safer', 'we use Claude so we're fine', 'our prompt is locked down'2m
02Direct injection vs indirect injection — the distinction that gets every team2m
03Real attack patterns — web-fetch payload, document poisoning, tool-output exfil, system-prompt extraction2m
04Agent → RCE — the Semantic Kernel pattern and why it generalizes to every agent framework2m
05Defenses Layer 1 — input sanitization + output validation + content filtering2m
06Defenses Layer 2 — tool-call allowlists, model-as-judge for safety checks2m
07Claude Code rules — ALLOWED_TOOLS, deny-by-default, sandbox mode1m
01'Denial of Wallet' — what Kevin called it, what OWASP folded into 'Unbounded Consumption'2m
02The actual cost: one unguarded /api/chat endpoint = unbounded Anthropic/OpenAI bill2m
03Why in-memory rate limiters DO NOT WORK on serverless — Vercel cold starts reset the Map2m
04Two-layer defense — Vercel Firewall (IP rate limit) + Supabase credits table (per-user daily budget)2m
05Monitoring + billing alerts — daily AI-spend checks, anomaly detection1m
01Row-Level Security — locked-by-default vs the demo mode that ships open2m
02The Supabase Advisors trick — Security + Performance, run before every deploy2m
03Service-role key NEVER in client code — how to spot it, how to fix it2m
04API keys — `.env.local` not `.env`, never `git add -A`, rotate when teammates leave2m
05Auth — magic links vs passwords vs SSO. httpOnly / secure / sameSite cookies2m
06Open ports, exposed admin URLs, leaked /api/_internal — the audit checklist2m
01SOC 2 — what it is, when you actually need it, the Vanta/Drata shortcut2m
02GDPR — data subject rights, deletion, sub-processors, the fines that keep landing1m
03HIPAA — PHI, BAAs, the three legal routes (AWS Bedrock+BAA, Anthropic Enterprise, local: Google MedGemma + Amazon HealthLake AI)2m
04ADA Title II Web Accessibility — original April 24, 2026 deadline EXTENDED to April 26, 2027 (state/local ≥50K) and April 26, 2028 (smaller). WCAG 2.1 AA mandatory. HHS May 2026 deadline still live.2m
05OWASP — what it is, why the LLM Top 10 is now mandatory reading2m
06License audit — MIT vs GPL/AGPL/SSPL. `npx license-checker --failOn` before every client delivery1m
01U.S. v. Heppner (Feb 2026, Judge Rakoff SDNY) — Claude chats NOT attorney-client privileged. The training-on-inputs problem2m
02Warner v. Gilbarco (E.D. Mich., Feb 2026) — the counter-ruling. Enterprise AI under counsel direction CAN be work-product protected1m
03The local-model trade-off — privilege/HIPAA win, but no harness like Claude Code. The honest math2m
04Claude with browser + computer access — what you're actually agreeing to. The trust budget2m
05Pen-test your own app WITH AI — Caido, PromptFoo for adversarial testing, garak2m
06Wrap-up: the 10 audit prompts you can run on your codebase tomorrow2m

What you take home

6 downloadable artifacts

Every course ships with the methodology AND the working assets. Drop them straight into your stack — they're AI-feedable, so you can paste them into Claude / Cursor / ChatGPT as a system prompt.

SKILL.md — AI security methodology

SKILL.md

The full prompt-injection + data-leakage + rate-limit defense playbook as an AI-feedable system prompt.

Vendor due-diligence checklist

Checklist

DPA + SOC 2 + GDPR + HIPAA + training-on-your-data — the questions every AI vendor must answer before you sign.

Contract clause library

Library

Pre-written contract clauses for both sides of the table — developer + business buyer.

10 AI audit prompts for your codebase

Prompt

Feed each to Claude / Cursor / ChatGPT to scan your repo for the specific vulnerabilities covered in the course.

API key storage matrix

Cheatsheet

.env.local / Vercel envs / GitHub Secrets / 1Password — what goes where, when, and why.

License audit one-liner

Code

The `npx license-checker --failOn` command to run before every client delivery — catches GPL/AGPL/SSPL.

Meet your instructor

Allen Seavert — in residence

Founder · SetupBots · Phoenix, AZ

13+ years building. Shipped marketplaces, CRMs, lead engines, and content pipelines for clients from solo founders to 8-figure companies. Built a production experiences marketplace from scratch — scraped 13,706 leads with AI, wired a seat-race booking system, and automated cold email at scale. Now teaching the methodology.

13+Years building
5Agent tools supported
100%Of the catalog — free

Frequently asked questions

No. This course is built for non-developers. Foundations like Claude Code, Supabase, Vercel, and GitHub are taught from zero. By the end you'll be shipping real apps without writing code yourself.
Lifetime access. You get every future update to the course as the tools evolve. No expiring subscriptions.
30-day no-questions-asked refund. If the course doesn't deliver on what we promised, we refund and we want to know why so we can improve.
Yes — completion certificates are issued automatically when you finish all course modules. They're shareable on LinkedIn.
Yes. Many operators buy a course first then book a consulting engagement to apply it to their specific business. Email growth@setupbots.com for details.

Other courses you may like

Hand-curated suggestions update over time based on what's most opened.

AI for Business Owners

20m·Free

Claude Code Fundamentals

2h 55m·Free

Getting Started with AI

22m·Free

AI Search — Get Cited by ChatGPT, Claude, Perplexity

2h 15m·Free

AI Web Design — The Flagship

7h 50m·Free

PostHog Setup (Analytics + Funnels + Replay)

4h 48m·Free
SUPER/SETUP · FREE IN EARLY ACCESS

The agent brain API.

Plug Claude Code, Codex, Cursor, VS Code, or Antigravity into the SetupBots brain — the methodology your agent reads on every build. Free with your account, like every course here.

  • 29+ courses, growing — all free
  • Brain API access for your agent
  • 25 hours of audio · drive, walk, lift
  • Community for you and your agent
Free with your account →
No card · generous daily allowance
Free

Secure account access


This course includes
  • 2h 35m on-demand video
  • 35 lessons · 6 modules
  • 6 downloadable artifacts
  • Access on web + mobile
  • Lifetime updates
Free2h 35m · 35 lessons
SetupBots

Agent intelligence, free starter resources, and implementation services for builders who want AI to ship safely.

support@setupbots.com · Phoenix, AZ

Resources

  • Free courses
  • SUPER/SETUP
  • Blog
  • Free tools

Services

  • Consulting
  • Done-for-you
  • Agent Intelligence API

Company

  • About
  • Contact
  • Events

Legal

  • Disclosure
  • Community Policy
  • Purchase Policy
© 2026 SetupBots. All rights reserved.Privacy·Terms·LicenseAll systems operational

Your cart · 0 items

Your cart is empty

Browse the catalog to add courses.

Browse courses
SetupBots
SUPER/SETUP
The agent brain API.
  • Free during early access — all of it
  • Your agent queries verified setup methodology
  • Claude Code, Codex, Cursor, VS Code, Antigravity
Free with your account
No card · generous daily allowance
Create free account Sign in