🌅 AI Daily Digest — April 16, 2026

Data Pulse

🚨 RED ALERT: Tennessee is about to make building chatbots a Class A felony — Tennessee is poised to enact legislation criminalizing chatbot development, classifying it as a Class A felony. The proposed law would impose 15 to 25 years in prison for creating or deploying such technology.
Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers — Objection, a Thiel-backed startup, uses AI to evaluate journalistic integrity and enable financial challenges to published stories. Critics argue the system could suppress whistleblowers by allowing users to dispute content.
Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference — Google’s Gemma 4 language model now runs natively on iPhones, enabling full offline AI inference without internet connectivity. This advancement allows users to leverage the model’s capabilities locally.
Hightouch reaches $100M ARR fueled by marketing tools powered by AI — Hightouch, a reverse ETL startup, has achieved $100 million in annual recurring revenue using AI-driven marketing tools. The company’s technology streamlines data integration for enterprise clients.
OpenAI updates its Agents SDK to help enterprises build safer, more capable agents — OpenAI has updated its Agents SDK to enhance enterprise development of safer, more capable agentic AI systems. The changes aim to improve safety and functionality for business applications.

BOAT: Navigating the Sea of In Silico Predictors for Antibody Design via Multi-O — Jackie Rao, Ferran Gonzalez Hernandez, Leon Gerard. Antibody lead optimization is inherently a multi-objective challenge in drug discovery.
Sandpile Economics: Theory, Identification, and Evidence — Diego Vallarino. Why do capitalist economies recurrently generate crises whose severity is disproportionate to the size of the triggering shock?
Forecasting Multivariate Time Series under Predictive Heterogeneity: A Validatio — Ziling Ma, Ángel López Oriona, Hernando Ombao. We study adaptive pooling under predictive heterogeneity in high-dimensional multivariate time series forecasting, where global models improve statistical efficiency but may fail to capture heterogene...
Online learning with noisy side observations — Tomáš Kocák, Gergely Neu, Michal Valko. We propose a new partial-observability model for online learning problems where the learner, besides its own loss, also observes some noisy feedback about the other actions, depending on the underlyin...
Spectral Thompson sampling — Tomas Kocak, Michal Valko, Remi Munos. Thompson Sampling (TS) has attracted a lot of interest due to its good empirical performance, in particular in the computational advertising.

New this week: