AI Risk Atlas Prototype/DemoUnofficial independent experiment. Not an official xAI product. Scores can be wrong.

Back to watch register
10Containment & agentsBelow the top 20

Escape via shared accelerators

CapabilityDomain knowledgeAffordanceImpact domainCap-adjacent

Statement (NASA form)

Given that training and inference tenants share GPUs, caches, and sometimes memory fabrics, there is a possibility of a tenant reading or influencing another tenant’s weights or prompts resulting in cross-customer leakage of secrets, and a path out of a logical sandbox.

Likelihood
2Remote
Consequence
4Critical
Urgency
2Planned

Side channels did not go away because the workload is a transformer.

Composite 10 = 2×4 + 2

Applicable mitigations

Related on the map