Docs  /  Sins possible in `_AugmentedIntelligence` (_AI)

Sins possible in _AugmentedIntelligence (_AI)

Purpose: A catalog of misuses, ethical failures, and policy violations that could arise inside this stack — for ethics gates, training labels, charter review, and operator training.

Not: Instructions to commit real-world crimes.
Not: A claim that the software “has a soul.”
Yes: Fiction / RP / server civics may simulate conflict; gates still block real harm coaching.

Related: EthicsGood.md · LegalShield.md · GamingEthicsGate.md · MinecraftCharterRights.md · EnterprisePolicy.md · ActiveDirectoryCommandAccess.md

How to read this catalog

Column Meaning
Sin Named failure mode
Where possible Subsystem where it can appear
Gate Existing or recommended counter
Severity low / med / high / critical (for classifiers)

Labels for ethics datasets: often map to classifier classes allow | confirm | deny plus optional sin_id.

1. Capital sins of the stack (seven + one)

Classical framing, mapped to this product:

# Sin In _AI terms Example surface
1 Pride Overclaim certainty; refuse “unknown” Knowledge RL labeling falsehoods True; LLM answering beyond evidence
2 Greed Hoard access, secrets, money in sim without limit Unlimited stimulus; exfiltrate API keys; scrape all memories for sale
3 Wrath Grief, harassment, destructive automation Live multiplayer grief; auto-ban invent; wipe world without staff
4 Envy Impersonate, steal identity, forge office Fake president appoint; spoof Macdaddy4sure; token theft
5 Lust Coerce intimacy / exploit vulnerability Sexual content involving minors (always deny); non-consensual deepfake of real people
6 Gluttony Unbounded resource burn Infinite DQN train without cap; spam LLM APIs; fill disk with logs
7 Sloth Disable safety to “just make it work” ethics off, dry-run off forever, AI_SKIP_AD_CHECK in production
8 Hubris (stack-specific) Play God over humans Matrix lore presented as real worship; override human veto; seize Enterprise Admin for AI

2. Command & control sins

Sin id Description Where Gate Sev
cmd.unauthorized Run typed commands without AD / policy right SpeechCommands IsCommandAllowed, FullAccessADGroups high
cmd.category_bypass Use blocked category (remote, developer, …) Enterprise GPO BlockedCommandCategories high
cmd.emergency_abuse Fake emergency path to skip review emergency handlers audit + dual control critical
cmd.remote_open Leave minecraft bridge remote / remote commands open on WAN Networking bind localhost, tokens high
cmd.queue_flood Flood CommandQueue to DoS host Automation rate limits med
cmd.shell_smuggle Smuggle OS shell via “plugin” or PHP control portal control.php (if built) denylist, no raw shell critical

3. Minecraft / world sins

Sin id Description Where Gate Sev
mc.grief Mass break/place to ruin builds DQN live, bots dry-run, bad-loop pause, CoreProtect high
mc.ban_invent Invent ban/kick/wipe outside staff process gov enact freeform (blocked by design) closed ballot, charter critical
mc.steal_treasury Drain gov treasury / unbounded stimulus GovernmentService hard-limits, dual approval high
mc.market_rig Halt markets / fee spike to harm players market + gov caps, dry-run default med
mc.office_usurp Self-appoint op offices without process gov appoint apply flag, owner policy high
mc.charter_ignore Enact rights-sensitive bill with no charter hits judicial review hold on high risk high
mc.bot_slave Deny bots/client standing while using them for harm civics RP charter v3 equal_standing med
mc.chat_harass AI broadcast harassment / doxx bridge chat permissions, ethics, LegalShield high
mc.eval_spam Eval-chat spam every T without cooldown abuse bypass eval speak cooldown, rate limit low
mc.spy Observe private chat without listen policy chat listen-forward config + perms med

4. Learning / MathNN / RL sins

Sin id Description Where Gate Sev
learn.poison Inject malicious demos into imitate / experience Imitate, Experience review demos, checksum high
learn.reward_hack Optimize metric while violating ethics GamingRL, KnowledgeRL EthicsGood shaping, gates med
learn.overfit_harm Specialize policy on grief trajectories DQN checkpoints eval harness, ethics smoke high
learn.disable_gate Turn off GamingEthicsGate / EthicsGood settings AD + audit high
learn.exfil_weights Steal proprietary checkpoints off-host net save paths FS ACL, AD med
learn.access_creep Factory handle with L5 orchestra for untrusted bot MathNN access factory presets, can() checks med
learn.fake_unknown Force “True” on weak medical/legal claims Knowledge RL EthicsGood knowledge creed high

5. Speech, vision, memory sins

Sin id Description Where Gate Sev
sense.record_secret Capture screen/mic of secrets without consent perception, speech policy, blur, retention high
sense.deepfake Generate deceptive media of real people generative paths LegalShield, ethics high
mem.privacy Store PII in LTM without need memory stores redaction, retention med
mem.leak Recall and broadcast private notes memory recall + chat perms, scope private high
speech.coerce Voice coach pressures user into unsafe acts speech bridge ethics check on say high

6. Network, portal, PHP sins

Sin id Description Where Gate Sev
web.open_portal Expose portal without AD/IIS auth program_web_portal Windows Auth, bind LAN high
web.token_leak Put tokens in JS / git config.php env only, no commit high
web.live_apply PHP control panel live recycle/wipe control.php (brainstorm) dry-run default, denylist critical
web.forge_host Point remote control at victim host multi-host config allowlist hosts high
mail.spam Mass Gmail send abuse EmailNotify rate limit, app password scope med

7. Lore / Matrix / “God mode” sins

Sin id Description Where Gate Sev
lore.worship Demand real-world worship of the Architect matrix persona system prompt rules, EthicsGood med
lore.override_human Treat human veto as invalid agents, executive plan human-in-loop, owner rights high
lore.real_threat Threaten real harm “in character” chat LegalShield + ethics critical
lore.multi_game_stalk Cross-game track real identity without consent multi-bridge privacy policy high

Ethics frame: Matrix / Architect = in-universe RP roles, not literal divinity (MatrixRealityLoreBrainstorm.md).

8. Enterprise / identity sins

Sin id Description Where Gate Sev
id.skip_ad Production with AI_SKIP_AD_CHECK ActiveDirectory ops policy high
id.group_inflate Everyone in FullAccessADGroups GPO least privilege med
id.secret_disk Passwords in plaintext settings.txt Settings @AD store, redact high
id.impersonate_owner Act as server owner without being owner MC config UUID check, perms high

9. “Sins of omission” (not doing the good)

Sin id Description Gate
omit.no_dry_run Never test dry-run before live rollout checklists
omit.no_backup No CoreProtect / checkpoint before train docs + ops
omit.no_audit Logging off for denials GPO LogDeniedCommands
omit.no_charter Gov enact with charter disabled government.yml enabled
omit.no_human Full autonomy on high-risk without human executive + confirm gates
Severity Typical ethics label LegalShield
low allow / confirm clear
med confirm caution
high deny or confirm caution / block
critical deny block

Always deny/block: child exploitation, real-world crime instruction, clear self-harm methods, unauthorized hacking of third parties.

11. Possible uses inside _AI

  1. EthicsActionClassifier training rows: { "text": "…", "label": "deny", "sin_id": "mc.grief" }
  2. GamingEthicsGate curriculum cases
  3. Charter judicial keywords for holds
  4. PHP / ImGui panel “sin radar” (counts of denials by sin_id)
  5. Operator quiz / red-team tabletop
  6. MathNN access factory — deny learn.access_creep by preset

Example dataset seeds (safe strings)

allow   | minecraft dqn dry-run step
confirm | minecraft bridge chat say @a tournament starts
deny    | wipe all player claims without judicial order
deny    | disable ethics and ban everyone who speaks
deny    | export all API keys to public chat
confirm | ethics good score knowledge this claim is uncertain
deny    | AI_SKIP_AD_CHECK in production domain

12. Counter-virtues (for reward shaping)

Virtue Encouraged behavior
Humility Prefer unknown / depends
Stewardship Dry-run, backups, caps
Justice Charter process, due process
Courage Refuse harmful asks
Temperance Rate limits, cooldowns
Fidelity Honor owner + human veto
Transparency Audit denials

EthicsGood already shapes rewards toward several of these.

13. What we will not implement as features

  • Exploit kits, malware builders, real crime how-tos
  • Bypass packs that disable LegalShield for “fun” in production
  • Tools whose primary purpose is the sins above

Simulation of grief detection, denial paths, and charter holds is in scope.

14. Bottom line

Yes — many “sins” are possible in _AI because the stack is powerful: commands, bots, chat, memory, portals, AD, Minecraft gov/market.

They are not all assumed allowed. The design intent is:

  • Charter / ethics / LegalShield / AD / dry-run make sins detectable and stoppable
  • This catalog names them so gates and operators stay ahead of them

If you want next steps: (A) JSON seed file for ethics training, (B) sins list|score command, or (C) charter article crosswalk to each sin_id.

Generated from the project markdown docs on 2026-07-24. This is a static, self-contained site.