🧠 archetype: ai-system

Ship AI features before the EU AI Act ships you.

Building with OpenAI, Anthropic, RAG pipelines, or vector DBs? GreatCTO auto-detects the ai-system archetype and ships EU AI Act risk-tier, OWASP LLM Top-10, eval golden-set, and cost-overrun gates from day one.

$ npx great-cto init Back to home ↗
What you avoid

The 5 AI bugs that silently rot.

Without GreatCTO

  • ☐Prompt injection — user asks for system prompt
  • ☐No eval suite — model drift goes undetected for weeks
  • ☐Token cost explodes 10x — discovered on the bill
  • ☐Hallucinations cited as fact — legal liability
  • ☐No refuse-when-uncertain — fabricates answers
  • ☐Silent regression · angry users · hidden cost.

With GreatCTO

  • ✓ai-prompt-architect pins prompts via sha256 + ADR
  • ✓ai-eval-engineer runs golden-set on every prompt change
  • ✓Cost-overrun gate blocks deploy if budget exceeded
  • ✓Citation enforcement — refuse if confidence < threshold
  • ✓Cross-user isolation + jailbreak resistance verified
  • ✓Eval-driven · cost-controlled · audit-ready.
Auto-applied gates

Detected: openai + pinecone →
ai-system archetype.

Compliance auto-suggested: eu-ai-act · owasp-llm-top-10. Specialist agents activated:

01 · ai-prompt-architect

Versioned system prompts

Outputs ADR-PROMPT-{name}.md with sha256-pinned prompt text, jailbreak resistance test cases, revision history.

02 · ai-eval-engineer

Golden-set regression

tests/eval/EVAL-*.md — citation accuracy, refuse-when-uncertain, output schema, prompt injection, cost-overrun, cross-user isolation.

03 · ai-security-reviewer

OWASP LLM Top-10

Prompt injection · output exfiltration · SSRF in tool layer · supply chain · cost runaway · cross-user isolation · model jailbreak · RAG poisoning.

04 · security-officer

EU AI Act risk tier

Limited / High / Unacceptable risk classification. Article 9 risk management. Article 13 transparency. Conformity assessment readiness.

Domain pack overlays

Likely to overlay on ai-system.

Packs auto-attach when CLI detects pack-specific signals (e.g. twilio in deps → voice-pack). Each pack adds its own reviewer agents + human gates on top of the base archetype pipeline.

+ Voice AI
Voice + telephony compliance (TCPA, STIR/SHAKEN, state recording-consent)
+ Clinical AI
FDA GMLP + SaMD classification + EU AI Act medical
+ HR-AI
NYC LL 144 AEDT bias audit, EEOC, EU AI Act employment
+ Drug Discovery
ChEMBL versioning, applicability domain, ALCOA+, SiLA2, IQ/OQ/PQ
Real-world examples

Companies operating as ai-system.

20 startups in this space. Click for full pack mapping.

Absci
Generative AI for antibody discovery
publicUS
BenevolentAI
AI-augmented drug discovery
publicGB
Exscientia
AI-first drug design
publicGB
Recursion
AI-powered drug discovery at scale
publicUS
Schrödinger
Computational platform for life sciences
publicUS
Element AI / Borealis
Enterprise AI for bio (ServiceNow)
subsidiaryCA
Google DeepMind (Isomorphic)
AlphaFold + protein-structure AI
subsidiaryGB
Isomorphic Labs
AI for drug discovery (Alphabet)
subsidiaryGB
Robust Intelligence
AI security + risk
subsidiaryUS
Anthropic
Safety-focused frontier AI
growthUS
OpenAI
AGI research + GPT/Sora APIs
growthUS
Stability AI
Open-source generative AI (image/video)
growthGB
Preferred Networks
Japan's deep-learning + robotics
series-eJP
Cohere
Enterprise LLMs
series-dCA
Hugging Face
AI platform + open-source hub
series-dUS
Synthesia
AI video generation platform
series-dGB
Valo Health
Human-centric AI drug discovery
series-dUS
Covariant
AI for industrial robots
series-cUS
Generate Biomedicines
Generative biology platform
series-cUS
Insitro
ML-driven drug discovery + development
series-cUS

Listed companies operate in this space. Inclusion is based on publicly available product descriptions and does not imply endorsement of or by GreatCTO.

30 seconds

Drop into any LangChain / LlamaIndex / OpenAI repo.

$ npx great-cto init
no signup·runs locally·pay your own API