""Your model is really valuable IP, and if you can distill the logic behind it, there's very real potential that you can replicate that technology - which is not inexpensive,""
""This is such an important technology, and the list of interested parties in replicating it are endless.""
""distillation attacks,""
""try to replicate Gemini's reasoning ability in non-English target languages across a wide variety of tasks.""
Google and OpenAI reported competitors, including DeepSeek, are probing models to extract underlying reasoning and copy capabilities. Threat actors include private-sector companies seeking valuable intellectual property from mature models. Google labeled the tactic "distillation attacks," and observed campaigns using over 100,000 prompts to try to replicate Gemini's reasoning across many tasks and languages. Abusing legitimate access to pretrained models can greatly reduce the cost and time required to develop new AI systems. Google says it detected a probe in real time and protected internal reasoning traces, but distillation remains a difficult risk to eliminate. Enforcement options include account blocking and legal action.
Read at Theregister
Unable to calculate read time
Collection
[
|
...
]