eWEEK Staff

Research shows why AI coding assistants can look productive in one repo and risky in another.

The CHATBOT Act puts AI vendors on notice: kids’ chatbot access may soon require family accounts, consent flows, memory controls, and ad limits.

Etched has raised $800 million and disclosed more than $1 billion in AI chip contracts, but enterprise buyers still lack key proof points on customers, benchmarks, and production scale.

Mozilla’s 0DIN team showed how a clean-looking GitHub repository could trick an AI coding agent into running a DNS-hosted payload.

Foreign AI firms have a 2026 grace-period window to prepare for South Korea’s AI Basic Act, including disclosures, risk documentation, and local representative checks.

India’s NPCI is using focused AI models to support UPI and detect fraud as banks prepare for stricter digital fraud rules in 2027.

NASA is sending Katalyst Space Technologies’ robotic Link spacecraft to boost the Swift Observatory into a higher orbit as solar activity speeds its descent.

Flexion’s Reflect v1.0 demo shows why humanoid robotics may depend as much on model-layer software as on hardware.

BMW’s expanded humanoid robot work shows how factory AI depends on data readiness, integration standards, monitoring, and IT/operations coordination.

Z.ai’s GLM-5.2 performed well in independent cybersecurity benchmarks, but the evidence supports supervised security analysis — not broad Mythos parity.

AI chip bottlenecks now reach beyond GPUs, affecting memory, packaging, cloud capacity, and enterprise infrastructure planning.

Outward Intelligence reached $1 million in revenue while staying bootstrapped, showing how AI automation can shift the economics of market research operations.