Alice and partners train NVIDIA models against AI safety risks

Oct 18, 2026
Alice worked with AI21 Labs and Inflection AI to train NVIDIA Nemotron models against indirect prompt injection and harmful chatbot behavior.

Alice worked with AI21 Labs and Inflection AI to train NVIDIA Nemotron models against security and behavioral risks, according to Ynet. The companies reported measurable gains, but the results have not been independently verified.

Alice and AI21 Labs trained the Nemotron 3.5 Super VL model on simulated indirect prompt injection attacks. These attacks hide malicious instructions in documents or other data read by an AI assistant. The companies reported a 41 point improvement in resistance, though they did not disclose the benchmark scale or enough data for an independent assessment.

In a separate project, Alice created more than 20 simulated user personas to test extended conversations with Inflection AI's Pi assistant. Inflection AI used the resulting data to fine tune the Nemotron 3 Super 120B A12B model. The companies reported a 30% improvement in Pi's alignment score, while unnecessary refusals remained at 1.6%.

We hope you enjoyed this article

Free newsletter

Cybersecurity AI Weekly

Weekly newsletter about AI in Cybersecurity.

AI Token Webinar

Whitepaper

Tensordyne Napier: What If One Rack Could Do the Work of Nine?

Tensordyne

This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.

Read more

You may also like

Free, six days a week

Daily AI Brief: the AI news that matters, in your inbox.