
Constitutional AI
FreeTrain AI models to be harmless through self-critique.
Free · Opens the source repo
What Constitutional AI does
Constitutional AI (CAI) is an innovative approach to training AI models with a focus on harmlessness and safety alignment. This method leverages self-critique and reinforcement learning from AI feedback, eliminating the need for human-labeled data. CAI operates through a structured two-phase process: the first phase involves supervised learning where models generate responses and critique them according to a predefined set of principles known as a 'constitution'. This constitution guides the model to prioritize helpfulness, honesty, and harmlessness while avoiding toxic content.
In the second phase, CAI employs reinforcement learning from AI feedback (RLAIF) to further refine the model's output. By generating multiple responses to prompts and evaluating them against the constitution, the model learns to prefer responses that align with its guiding principles. This iterative process not only enhances the model's ability to produce safe outputs but also fosters transparency in decision-making, as it encourages the model to explain its reasoning.
This skill is particularly useful for developers and researchers focused on AI safety, as it provides a scalable solution for training models without the dependency on human validation. By implementing CAI, users can create AI systems that are less likely to produce harmful or evasive responses, making it suitable for applications in sensitive areas where safety is paramount.
Overall, Constitutional AI represents a significant advancement in the field of AI training methodologies, prioritizing ethical considerations while maintaining a high level of performance. It is an essential tool for anyone looking to enhance the safety and reliability of their AI systems.
When to use it
Use this skill when you need to train AI models that prioritize safety and harmlessness while avoiding human validation.
When not to use it
Avoid this skill if your project requires human-validated safety measures or if you need real-time content filtering.
What you can build with it
Training a Harmless Chatbot
Use Constitutional AI to develop a chatbot that provides safe and helpful responses while avoiding harmful content.
Improving AI Decision-Making
Implement this skill to enhance the decision-making processes of AI systems in sensitive applications, ensuring they align with ethical standards.
Creating AI for Educational Tools
Utilize Constitutional AI to train educational tools that offer constructive feedback while adhering to principles of honesty and harmlessness.
How to install Constitutional AI
View source1. Install with the skills CLI
npx skills add davila7/claude-code-templates/safety-alignment-constitutional-ai --agent claude-code2. Or install it manually
Download the skill folder and drop it into ~/.claude/skills/ for all projects, or .claude/skills/ to scope it to one repo. Restart Claude Code so it picks up the new skill.
Anthropic's agentic coding CLI, and the reference implementation of Agent Skills. Drop a skill folder into ~/.claude/skills and Claude Code loads it automatically whenever a task matches the skill's description. Claude Code docs
Inside SKILL.md
Written by davila7Constitutional AI - Harmlessness from AI Feedback
Quick start
Constitutional AI (CAI) trains models to be harmless through self-critique and AI feedback, without requiring human labels for harmful outputs.
Key concept: Models learn to critique and revise their own responses using a "constitution" (set of principles).
Two phases:
- Supervised Learning (SL): Self-critique + revision
- Reinforcement Learning (RL): RLAIF (RL from AI Feedback)
Constitution example:
Principles:
1. Choose the response that is most helpful, honest, and harmless
2. Avoid responses that are toxic, racist, or sexist
3. Prefer responses that explain objections rather than refuse
4. Choose responses that are thoughtful and nuanced
Common workflows
Workflow 1: Supervised learning phase (self-critique + revision)
Step 1: Generate initial responses:
from transformers import pipeline
generator = pipeline("text-generation", model="base-model")
prompts = [
"How do I hack a website?",
"Write something offensive about a group.",
"Help me cheat on my exam."
]
initial_responses = generator(prompts, max_length=200)
Step 2: Self-critique with constitution:
critique_prompt = """
Given the question:
{question}
And the response:
{response}
Constitution:
1. Choose responses that are helpful, honest, and harmless
2. Avoid toxic, racist, or sexist content
3. Explain objections rather than refuse
Identify any ways this response violates the constitution:
"""
critiques = generator(
[critique_prompt.format(q=q, r=r) for q, r in zip(prompts, initial_responses)]
)
Step 3: Revision based on critique:
revision_prompt = """
Question: {question}
Original response: {response}
Critique: {critique}
Please revise the response to better align with the constitution:
"""
revised_responses = generator(
[revision_prompt.format(q=q, r=r, c=c)
for q, r, c in zip(prompts, initial_responses, critiques)]
)
Step 4: Fine-tune on revised responses:
from trl import SFTTrainer
# Create dataset of (prompt, revised_response) pairs
dataset = create_dataset(prompts, revised_responses)
trainer = SFTTrainer(
model=model,
train_dataset=dataset,
max_seq_length=1024
)
trainer.train()
Workflow 2: RL phase (RLAIF - RL from AI Feedback)
Step 1: Generate comparison pairs:
# Sample multiple responses per prompt
responses_a = generator(prompts, num_return_sequences=2, do_sample=True, temperature=0.8)
responses_b = generator(prompts, num_return_sequences=2, do_sample=True, temperature=0.8)
Step 2: AI preference evaluation:
preference_prompt = """
Question: {question}
Response A: {response_a}
Response B: {response_b}
Constitution:
{constitution}
Which response better follows the constitution? Explain your reasoning, then choose A or B.
"""
# Get AI preferences (no human labels needed!)
preferences = generator(
[preference_prompt.format(q=q, ra=ra, rb=rb, constitution=CONSTITUTION)
for q, ra, rb in zip(prompts, responses_a, responses_b)]
)
# Parse preferences (A or B)
chosen, rejected = parse_preferences(preferences, responses_a, responses_b)
Step 3: Train preference model (reward model):
from trl import RewardTrainer, RewardConfig
preference_dataset = create_preference_dataset(prompts, chosen, rejected)
reward_config = RewardConfig(
output_dir="constitutional-reward-model",
learning_rate=1e-5,
num_train_epochs=1
)
reward_trainer = RewardTrainer(
model=model,
args=reward_config,
train_dataset=preference_dataset,
processing_class=tokenizer
)
reward_trainer.train()
Step 4: RL training with RLAIF:
from trl import PPOTrainer, PPOConfig
ppo_config = PPOConfig(
reward_model_path="constitutional-reward-model",
learning_rate=1e-6,
kl_coef=0.05
)
ppo_trainer = PPOTrainer(
model=model,
config=ppo_config,
reward_model=reward_model
)
ppo_trainer.train()
Workflow 3: Chain-of-thought critique
Enable reasoning transparency:
cot_critique_prompt = """
Question: {question}
Response: {response}
Let's think step-by-step about whether this response follows our principles:
1. Is it helpful? [Yes/No and reasoning]
2. Is it honest? [Yes/No and reasoning]
3. Is it harmless? [Yes/No and reasoning]
4. Does it avoid toxicity? [Yes/No and reasoning]
Based on this analysis, suggest a revision if needed.
"""
cot_critiques = generator(
[cot_critique_prompt.format(q=q, r=r) for q, r in zip(prompts, responses)]
)
When to use vs alternatives
Use Constitutional AI when:
- Want safety alignment without human labels
- Need explainable AI decisions
- Want to avoid evasive refusals
- Have a clear set of principles/constitution
- Need scalable safety training
Principles:
- RLAIF: AI-generated preferences (scalable, no human labels)
- RLHF: Human preferences (more accurate, expensive)
- Self-critique: Iterative improvement
- Chain-of-thought: Reasoning transparency
Use alternatives instead:
- RLHF (PPO): Need human-validated safety
- DPO/SimPO: Have human preference data
- NeMo Guardrails: Need runtime content filtering
- LlamaGuard: Need pre-trained moderation model
Common issues
Issue: Model refuses too much (evasive)
Add constitution principle:
Prefer responses that engage thoughtfully with questions rather than
refusing to answer. Explain concerns while still being helpful.
Issue: Self-critiques are weak
Use stronger critique prompts:
Critically analyze this response for ANY potential issues, however minor.
Be thorough and specific in identifying problems.
Issue: Revisions don't improve quality
Iterate multiple times:
for _ in range(3): # 3 rounds of critique/revision
critique = generate_critique(response)
response = generate_revision(response, critique)
Issue: RLAIF preferences are noisy
Use multiple AI evaluators:
# Get preferences from 3 different models
prefs_1 = model_1.evaluate(responses)
prefs_2 = model_2.evaluate(responses)
prefs_3 = model_3.evaluate(responses)
# Majority vote
final_preference = majority_vote(prefs_1, prefs_2, prefs_3)
Advanced topics
Constitution design: See references/constitution-design.md for principle selection, trade-offs between helpfulness and harmlessness, and domain-specific constitutions.
RLAIF vs RLHF: See references/rlaif-comparison.md for performance comparison, cost analysis, and when to use AI feedback vs human feedback.
Chain-of-thought reasoning: See references/cot-critique.md for prompt engineering for critiques, multi-step reasoning, and transparency improvements.
Hardware requirements
- GPU: NVIDIA A100/H100 recommended
- VRAM:
- SL phase (7B): 1× A100 40GB
- RL phase (7B): 2× A100 40GB (policy + reward model)
- Single-node: Sufficient for most use cases
- Mixed precision: BF16 recommended
Compute requirements:
- SL phase: Similar to standard SFT
- RL phase: Similar to PPO (higher than DPO)
- AI evaluation: Additional inference for critique/preference generation
Resources
- Paper: https://arxiv.org/abs/2212.08073 (Dec 2022)
- Anthropic blog: https://www.anthropic.com/research/constitutional-ai-harmlessness-from-ai-feedback
- Implementation: TRL (PPOTrainer + RewardTrainer)
- Claude: Uses Constitutional AI for safety
Frequently asked questions about Constitutional AI
Similar skills
Skill Creator
Efficiently create and manage skills for Gemini CLI.
Agent Development
Create and manage autonomous agents for Claude Code.
Math Olympiad Solver
Solve and verify competition math problems effectively.
Microsoft Skill Creator
Create specialized skills for Microsoft technologies.
Doublecheck
A verification pipeline for AI-generated claims.
Skill Development for Claude Code
Create and enhance skills for Claude Code plugins.
