Safety best practices

Developing AI Systems with the OpenAI API

Francesca Donadoni

Curriculum Manager, DataCamp

Safety with the OpenAI API

A circle with the OpenAI API

Safety with the OpenAI API

A circle with the OpenAI API that intersects with another labeled 'ethics and fairness'

Safety with the OpenAI API

A circle with the OpenAI API that intersects with other two labeled 'ethics and fairness' and 'alignment'

Safety with the OpenAI API

A circle with the OpenAI API that intersects with other three labeled 'ethics and fairness', 'alignment' and 'privacy of data'

Safety with the OpenAI API

A circle with the OpenAI API that intersects with other four labeled 'ethics and fairness', 'alignment', 'privacy of data', and 'security of the system'

Best practices

Moderation API
Adversarial testing
Limit user input and output tokens
Prompt engineering

¹ https://platform.openai.com/docs/guides/safety-best-practices

Best practices

Human in the loop
"Know your customer"
Allow users to report issues

An illustration of customer and developer working together to improve the product based on feedback

Best practices

Keep your API keys safe
Communicate limitations

A shield representing safety

Speech bubbles representing communication

Using end-user IDs

import uuid

unique_id = str(uuid.uuid4())

response = client.chat.completions.create(  
  model="gpt-4o-mini", 
  messages=messages,
  user=unique_id)

print(unique_id)

ad58af43-19b1-4fd8-9905-5ed3bce7e008

Keeping your API key safe

Unique API key
Server-side security
Avoid repository exposure
Environment variables
Key management services
Monitor and rotate keys

Let's practice!

Developing AI Systems with the OpenAI API