Phishing is the practice of tricking people into revealing sensitive information, and has long been a top cyber threat. But in 2025, phishing has entered a new era: attackers are now using AI tools to generate highly convincing and customized phishing messages at scale.

AI chatbots have become a daily fixture, from customer support and virtual assistants to AI companions. But behind their smooth conversations lies a complex system that even their creators sometimes struggle to fully explain. This opacity is what many refer to as the “black box” problem in AI: we see the inputs and outputs, but not always the logic in between.

AI is transforming consumer lending, promising faster credit decisions and greater access. But the reality is more complicated. Many AI-based scoring systems rely on alternative data — utility payments, online behavior, even social media activity — to predict creditworthiness.

In early 2025, TikTok quietly updated its privacy policy for users in the U.S. and Europe. The changes, buried deep in the fine print, outline broader data-sharing arrangements with its parent company ByteDance, as well as unnamed “business partners.” Privacy advocates warn that these updates could expose users to greater tracking, profiling, and potential government access.

Cybersecurity experts are warning about WormGPT, a black-market AI tool designed for criminals. Unlike mainstream AI models that include content filters, WormGPT deliberately omits safeguards, allowing users to generate phishing emails, malware code, and scam scripts at scale.

Almost every website now greets visitors with a cookie consent banner, claiming to offer a choice over tracking. But privacy advocates warn that many of these pop-ups are designed not to inform, but to manipulate. Known as dark patterns, these interfaces use visual tricks to steer users toward accepting all cookies, often burying “reject all” options behind extra clicks, small text, or confusing layouts.

A new IBM global study highlights a troubling gap between AI policy rhetoric and real-world governance. Although 80% of executives say AI governance is critical to responsible innovation, only 20-30% have put robust practices in place.

New Orleans is at the center of a national debate as it considers an ordinance to legalize real-time facial recognition surveillance across the city.

Europe’s landmark AI Act has officially entered the implementation phase, ushering in one of the world’s most comprehensive frameworks for governing artificial intelligence. The law targets high-risk AI systems, including biometric identification, predictive policing, and AI in critical infrastructure, with strict requirements for transparency, human oversight, and risk assessments.