Google TechTalks

Google TechTalks

Differential PrivacyMachine LearningData PrivacyLarge Language Models (LLMs)Machine Learning SecurityData poisoningData SecurityPrompt EngineeringFine-tuningLarge Language ModelsPrivacy AuditingLLM securityFederated LearningAI EthicsAdversarial AttacksMembership Inference AttacksModel MemorizationDeep LearningMachine learning vulnerabilitiesSynthetic Data GenerationMachine Learning PrivacyRetrieval Augmented Generation (RAG)AI SecurityNatural Language ProcessingLanguage ModelsAI SafetyContinual CountingGenerative AIStreaming AlgorithmsApproximation AlgorithmsData MemorizationPrivacyPrivacy-Preserving Data AnalysisCopyright InfringementInformation Theory
Cascading Adversarial Bias from Injection to Distillation in Language Models
Language ModelsAdversarial AttacksData Poisoning

Cascading Adversarial Bias from Injection to Distillation in Language Models

Adversarial bias injected into large language models (LLMs) during instruction tuning can cascade and amplify in distilled student models, even with minimal poisoning, bypassing current detection methods.

Explore Insights →
Persistent Pre-Training Poisoning of LLMs
LLM securityAI safetyData poisoning

Persistent Pre-Training Poisoning of LLMs

Adversaries can persistently compromise Large Language Models (LLMs) by injecting a small amount of malicious data (as little as 10 tokens per million) into their pre-training datasets, leading to behaviors like denial of service, private data extraction, and belief manipulation, even after subsequent alignment training.

Explore Insights →

Want more on ai safety?

Explore deep-dive summaries and actionable takeaways from the best minds across different podcasts discussing this topic.

View All Ai Safety Episodes→

Don't see the episode you're looking for?

We're constantly adding new episodes, but if you want to see a specific one from Google TechTalks summarized, let us know!

Submit an Episode