Microsoft Research

New research from Microsoft Research demonstrates that reprogramming just a few dormant neurons in an AI agent can create a backdoor that leaves the agent behaving normally until it hits a specific trigger.

While the specifics of this threat are novel, the mitigation is not: Know where your model came from, and be smart about its permissions. Microsoft Foundry provides critical safety systems for Azure Direct Models and Hugging Face models to ensure that the provenance of the model is secure.

Microsoft researcher @DimitrisPapail explains the measures that model hubs and vendors should take as well as mitigation strategies for enterprises and users. msft.it/6014ahp6e

How do you test Copilot against real-world business complexity without exposing customer data? Microsoft researchers created a platform that generates synthetic enterprise environments, complete with simulated employees, meetings, emails, documents, and industry-specific workflows. These environments let Microsoft evaluate Copilot at scale in realistic business scenarios while preserving customer data privacy.

Read about the Industry Evaluation Data Generation Platform: msft.it/6011ahGkZ

Lilika Makabe’s internship provided meaningful connections and opportunities to grow; can AI bring us Star Trek's universal translator; solving a decades-old math challenge; smarter context compaction boosts coding agents; and building a better tokenizer: msft.it/6019aCuhH

What does it take to grow more confident in your future as a researcher?

For Tony Chae, an undergraduate research internship at Microsoft Research offered an opportunity to explore new ideas, collaborate with researchers, and build connections beyond the lab.

The experience helped Tony expand his knowledge, develop new perspectives, and build confidence as he prepares for a future PhD.

Every technological revolution has remodeled our institutions – how might AI?
In @ScienceMagazine, Microsoft researcher @glenweyl and co-author @profjamesevans explore ‘reverse alignment,’ the idea that using neural networks as a model for reshaping institutions might offer a path toward harnessing the benefits of AI while managing its risks. msft.it/6015aAHYP