GPT-2 Model Replication by External Researchers

# GPT-2 Model Replication by External Researchers

https://gcdradar.vercel.app/incidents/gpt-2-model-replication-by-external-researchers · AIR-2020-0024 · closed 2020-11-06

- Lane: Research
- Severity: 1 (Research / proof of concept)
- Fix status: Fix status unknown
- First reported: 2020-10-23 (happened 2019-08-22)
- AI role: Attack on AI
- Products: GPT-2

## Summary

Researchers successfully replicated OpenAI's GPT-2 language model prior to its full release using publicly available technical information. The event highlighted risks related to tiered release schedules.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2020-10-23: Medium BlogPost, "OpenGPT-2: We Replicated GPT-2 Because You Can Too" (https://blog.usejournal.com/opengpt-2-we-replicated-gpt-2-because-you-can-too-45e34e6d36dc)
- 2020-10-23: Wired Article, "OpenAI Said Its Code Was Risky. Two Grads Re-Created It Anyway" (https://wired.com/story/dangerous-ai-open-source)
- 2020-10-23: GPT-2 Model Replication (https://atlas.mitre.org/studies/AML.CS0007)

## Sources

- 2020-10-23 MITRE ATLAS: GPT-2 Model Replication https://atlas.mitre.org/studies/AML.CS0007
- 2020-10-23 Wired: Wired Article, "OpenAI Said Its Code Was Risky. Two Grads Re-Created It Anyway" https://wired.com/story/dangerous-ai-open-source
- 2020-10-23 blog.usejournal.com: Medium BlogPost, "OpenGPT-2: We Replicated GPT-2 Because You Can Too" https://blog.usejournal.com/opengpt-2-we-replicated-gpt-2-because-you-can-too-45e34e6d36dc

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.