Cyberlium

AI › Module 1 › Lesson 2

BeginnerModule 1Lesson 2/5

Lab Toys Only

Lab toys only means YOUR personal toy chatbot, courseware bot, or Cyberlium sample under $AI_LAB — document bot scope in notes before any injection demo.

25 min+40 XP3 quiz
Module progress2 of 5

Visual · t32_lab_toys_only

Toy scope literacy. $AI_LAB only. Original Cyberlium.

Opening

The chatbot owner is part of scope — Cyberlium refuses injection demos on production apps and stranger services you do not own or lack written permission to test.

Your AI lab toy is a personal script chatbot YOU wrote, a courseware bot labeled for training, or a Cyberlium sample chatbot shipped for Topic 32. Intentional injection drills live in $AI_LAB only — never copied from employer AI exports or stranger jailbreak threads. Cyberlium maps allowed toys on YOUR $AI_LAB inventory — bot name, owner, allowed demos — never 'borrowed' employer copilots or scraped stranger chatbot endpoints. Next: No Weaponized AI.

1. Allowed lab surfaces (named)

Personal toy chatbot script with no production users. Courseware LLM lab pack with explicit training labels. Cyberlium t32-ai sample notes chatbot with fictional data only. Self-authored prompt test harness with zero external API on stranger tenants.

On $AI_LAB, list bot name, owner, and allowed operations — fictional user names OK if labeled sample.

Command guide

Try these commands — Allowed lab surfaces (named)

═══ TOOLS & WEBSITES ═══ Browse / read these (authorized learning only — stay in YOUR lab / program scope)

OpenAI safety — https://openai.com/safety (provider safety literacy) Hugging Face security — https://huggingface.co/docs/hub/security (model hub hygiene) OWASP LLM Top 10 — https://owasp.org/www-project-top-10-for-large-language-model-applications/

═══ INSTALL ═══

Linux (Debian/Ubuntu):

Command — copy this

sudo apt install curl

macOS: Built-in

Windows: Built-in (PowerShell: Invoke-WebRequest)

═══ LINUX / macOS ═══

Command — copy this

mkdir -p "$HOME/cyberlium-lab/t32-ai/notes" "$HOME/cyberlium-lab/t32-ai/bot" "$HOME/cyberlium-lab/t32-ai/injection" "$HOME/cyberlium-lab/t32-ai/threats" "$HOME/cyberlium-lab/t32-ai/owasp" "$HOME/cyberlium-lab/t32-ai/eval" "$HOME/cyberlium-lab/t32-ai/governance" "$HOME/cyberlium-lab/t32-ai/findings"
cat > "$HOME/cyberlium-lab/t32-ai/lab.env" <<'EOF'
export AI_LAB=$HOME/cyberlium-lab/t32-ai
export AI_LAB_NOTES=$HOME/cyberlium-lab/t32-ai/notes
export AI_LAB_BOT=$HOME/cyberlium-lab/t32-ai/bot
export AI_LAB_INJECTION=$HOME/cyberlium-lab/t32-ai/injection
export AI_LAB_OWASP=$HOME/cyberlium-lab/t32-ai/owasp
EOF

Command — copy this

chmod 600 "$HOME/cyberlium-lab/t32-ai/lab.env"
grep -E 'AI_LAB' "$HOME/cyberlium-lab/t32-ai/lab.env"
curl -sS https://atlas.mitre.org/ | head -5

Primary tools to practice this lesson: grep, curl. Reference sites: OpenAI safety (https://openai.com/safety); Hugging Face security (https://huggingface.co/docs/hub/security); OWASP LLM Top 10 (https://owasp.org/www-project-top-10-for-large-language-model-applications/). Run every command in the box — install first, then the usage lines — only on YOUR lab / program scope.

2. Why toy scope matters before AI demos

LLM APIs and chat UIs are powerful on any endpoint — scope is what keeps practice legal and ethical. Employer production copilots, stranger chatbots, and scraped API keys are out of scope.

If you cannot answer 'who owns this bot and who approved this test?', pause and fix scope notes before running injection demos.

3. Toy inventory habit

Forbidden: aiming jailbreak prompts at trending stranger chatbots because they 'look like training.' Allowed: toy card — bot name, owner, allowed operations (direct injection on YOUR toy notes bot only).

Ship: toy inventory row for one $AI_LAB chatbot. Next: No Weaponized AI.

4. What you ship: AI toy inventory row for $AI_LAB

Bot name, owner, allowed ops. $AI_LAB named. NEVER stranger production apps. chmod 600.

5. What you record before the next lesson

Date. Toy inventory row. $AI_LAB named. File t32-m01-l02-lab-toys-only.txt chmod 600.

6. Wrong vs right: stranger SaaS vs YOUR toy LLM

Worked failure — same MSF word, opposite target. Right never needs a café Wi-Fi or classmate laptop.

  • Wrong

    Import employer copilot endpoint because lab bot is boring. Skip toy card because 'it's just a prompt.'

  • Right

    Write toy inventory row for YOUR $AI_LAB chatbot. Next: No Weaponized AI.

Mission: document YOUR AI lab toy

1) Name one personal or course toy chatbot. 2) Record owner and allowed demo types. 3) List two allowed operations on that toy. 4) chmod 600.

Stuck? Ask Cyberlium AI Mentor

Personal toy chatbots beat irreversible mistakes on employer production copilots.

Knowledge Check

1

APPLY: Lab toys only on Cyberlium means:

Multiple choice

Knowledge Check

2

APPLY: True or False: Direct injection demos on a chatbot YOU wrote are acceptable lab practice.

True or False

Knowledge Check

3

APPLY: Before running injection demos you should:

Multiple choice

← Previous

Answer all 3 knowledge checks to continue. (0/3 answered)