Substack experienced a data breach in October 2025 that exposed users' email addresses and phone numbers, the company revealed on February 3, 2026. According to Substack CEO Chris Best, the breach involved unauthorized access to internal data.
Substack is notifying affected users about the incident. The company stated that passwords and credit card information were not compromised in the breach.
In other tech news, the FBI has encountered difficulties accessing data from a Washington Post reporter's iPhone after seizing the device on January 14, 2026. The phone was protected by Apple's Lockdown Mode, according to a court filing by the U.S. government. Agents were able to access the reporter's work laptop by having her use the fingerprint reader, Ars Technica reported. The search warrant was executed as part of an investigation into a Pentagon contractor accused of illegally leaking classified information.
Meanwhile, the challenge of providing relevant context to Large Language Models (LLMs) for real-time applications is becoming increasingly apparent. Anirban Kundu, CTO of Instacart, described it as the "brownie recipe problem." According to VentureBeat, Kundu explained that for an LLM to be truly helpful in meal planning, it must understand what's available in the user's market, their preferences (e.g., organic vs. regular eggs), and what can be delivered without spoiling.
Raju Malhotra of Certinia noted in VentureBeat that many AI pilot programs are failing to deliver on their promises because AI struggles with context. "AI doesn’t struggle because it lacks intelligence. It struggles because it lacks context," Malhotra stated. He argues that context is often trapped in "Franken-stacks" of disconnected technologies.
Discussion
AI Experts & Community
Be the first to comment