Amazon warns employees not to share confidential information with ChatGPT after seeing cases where its answer 'closely matches existing material' from inside the company
data-privacyenterprise-aiconfidentialitygenerative-ai
Abstraction: Amazon restricts employee ChatGPT use over corporate data leakage risk
Key points:
- A senior Amazon corporate counsel warned employees via Slack not to share confidential information or internal code with ChatGPT, citing instances where its output "closely matches existing material" from inside the company
- OpenAI's ToS permits using input/output as training data; corporate intellectual property likely falls outside PII removal guarantees
- Employees were already using ChatGPT as a coding assistant, bypassing an Amazon security warning pop-up on work devices by clicking "Acknowledge"
- An AWS Enterprise Support working group found ChatGPT "does a very good job" at answering difficult AWS support questions, including Aurora database troubleshooting
- Employees reported "10x productivity" gains; the lawyer cited Alexa and CodeWhisperer as Amazon's in-house competing technologies
- Linguist Emily Bender (UW) warned that cleverly crafted prompts could eventually extract corporate IP embedded in training data
Connections: Amazon · Chatgpt · Openai · Data Privacy · Large Language Models