Skip to content
OpenAI Disrupts Model-Copying Campaign Linked to Moonshot AI

OpenAI Disrupts Model-Copying Campaign Linked to Moonshot AI

First seen 1 Oct 2026, 19:05 UTC • •

Article Content

Browse articles
ThreatCluster AI
ThreatCluster •October 2, 2026 at 15:08 UTC
  • •OpenAI disrupted a campaign to extract reasoning from its AI models linked to Moonshot AI.
  • •The activity peaked in late July with 16,000 requests from over 4,000 users.
  • •No encryption breach occurred; attackers manipulated model interactions instead.

OpenAI has disrupted a coordinated campaign aimed at extracting hidden reasoning from its AI models, attributing the activity to Chinese startup Moonshot AI, developer of the Kimi chatbot. The campaign, characterized as 'adversarial distillation,' began on July 1 and peaked with 16,000 requests from over 4,000 users in late July. OpenAI confirmed that the attackers did not breach encryption or access stored user conversations but manipulated model interactions to reveal protected reasoning. The company identified a network of more than 15,000 accounts involved in the attempts, which were fully shut down by July 28. OpenAI has since implemented countermeasures, including banning fraudulent accounts and tightening sign-up processes. The findings were shared with other AI developers and government channels to address broader implications beyond OpenAI's models.

Start a free Starter trial for enhanced analysis

Ask AI about this cluster

Updated 44d ago How this analysis works

Timeline

2026-07-01
Campaign begins
OpenAI detected initial low-volume attempts to extract reasoning from its models.
The-Decoder
2026-07-24
Surge in activity
The campaign surged to 16,000 requests from over 4,000 users over two days.
Cnbc
2026-07-28
Campaign fully disrupted
OpenAI reported that it had fully shut down the network of fraudulent accounts involved in the campaign.
Heygotrade

More articles in this cluster (7)

Following this threat?

Track Anthropic in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.

Free account · no card needed

Common questions

What specific methods were used in the attack?
The attackers manipulated model interactions to extract reasoning without breaching encryption.
What actions has OpenAI taken in response?
OpenAI has banned fraudulent accounts, tightened sign-up processes, and implemented countermeasures to protect its models.
Is there a risk of similar attacks on other AI models?
Yes, OpenAI shared its findings with other AI developers and government channels to address potential risks across the industry.