AI News for April 6, 2026 — 7min.ai

[NEWS] Suno's copyright filters are 'incredibly easy to fool,' generating near-copies of hit songs: The Verge found that Suno's AI music platform produces alarmingly close imitations of copyrighted songs like Beyoncé's "Freedom" and Black Sabbath's "Paranoid" with minimal effort.
[RESEARCH] AI offensive cyber capabilities doubling every 5.7 months, Lyptus Research finds: AI safety firm Lyptus Research tested AI models against 291 cybersecurity tasks with 10 professional security experts.
[NEWS] Chinese AI firms race to capture OpenClaw users displaced by Anthropic's exit: Chinese tech companies are publicly competing to attract developers and users left without access after Anthropic restricted Claude models from OpenClaw.
[NEWS] Grammarly's 'Expert Review' impersonated real experts with AI-generated suggestions under their names: The Verge detailed Grammarly's (now rebranded as Superhuman) controversial "Expert Review" feature that generated AI suggestions under real experts' names and faces without their consent.
[NEWS] 'FOBO' (Fear of Becoming Obsolete) doubles as AI anxiety reshapes American workforce: Four in 10 workers now cite AI-driven job loss as a primary fear, nearly double the rate from a year ago, per KPMG.
[NEWS] Japan targets 30% of global physical AI market by 2040 as labor shortages drive robot deployment: Japan's Ministry of Economy aims to capture 30% of the global physical AI market by 2040, building on its existing 70% share of industrial robotics.
[OPINION] Iran war's energy disruption threatens the fragile economics of AI data centers: Beyond the broad AI slowdown, The Guardian identifies AI data centers as uniquely exposed to the Iran war's energy price shock: they are energy-hungry, operate on unproven business models, and carry massive debt that assumed stable power costs.
[RESEARCH] Google Research: AI benchmarks need at least 10 human raters per example, not the standard 3-5: A study by Google Research and Rochester Institute of Technology found that standard AI benchmarks using 3-5 human evaluators per test example produce unreliable results.
[NEWS] AI data center boom 'stress tests' insurers as private capital floods in: The surge in AI data center construction is straining the insurance industry, with insurers scrambling to assess risks for GPU-backed debt deals and massive infrastructure projects.
[OPINION] Mark Cuban warns CEOs face lose-lose 'Innovator's AI Dilemma': Mark Cuban says CEOs of public companies face an impossible choice: tear down and rebuild as AI-native, or do nothing.

About 7min.ai · Subscribe to newsletter