AWS Addresses Security Concerns with OpenAI's ChatGPT at re:Invent
During AWS's re:Invent event, AWS's chief, Adam Selipsky, subtly addressed security concerns related to OpenAI's ChatGPT. In response, AWS introduced safety features called Guardrails in Amazon Bedrock.
Responsible AI Integration
Selipsky emphasized the importance of responsible AI and how AWS has integrated it into their platform. Responsible AI involves ensuring safe interactions between users and applications, preventing harmful outcomes. One key approach is setting limits on what information models can and cannot do.
Introducing Guardrails for Amazon Bedrock
Guardrails for Amazon Bedrock enable users to implement consistent safeguards, ensuring relevant and safe user experiences aligned with company policies. Key features include:
- Setting restrictions on topics
- Applying content filters to eliminate undesirable and harmful content in interactions within applications
Applicability to Language Models in Amazon Bedrock
These guardrails can be applied to all language models (LLMs) in Amazon Bedrock, including fine-tuned models and Agents for Amazon Bedrock.
OpenAI's Perspective on Safety
Greg Brockman, OpenAI's former board member, commented on AWS's Guardrails, highlighting OpenAI's unique perspective on safety. OpenAI emphasizes safety through scientific measurement and lessons from iterative deployment. OpenAI also assures users that API data is not used to train their models.
Building Trust with ChatGPT Enterprise
In an effort to build trust with enterprises, OpenAI introduced ChatGPT Enterprise earlier this year.