AI Economics, Security Gaps, and Chip Advances

AI Economics, Security Gaps, and Chip Advances

OpenAI Chairman Explains AI Tokenomics

OpenAI Chairman Explains AI Tokenomics

OpenAI Chairman Bret Taylor appeared on Squawk Box to address the AI boom's economics, covering token efficiency, ROI on AI spending, the competitive landscape, and an Apple lawsuit. The discussion centers on "tokenmaxxing"—how AI systems consume and optimize tokens—a key factor shaping costs and returns for businesses deploying AI at scale.

via cnbc.com · also covered by cnbc.com
Guardrails Blocked Defenders, Not Attackers

Guardrails Blocked Defenders, Not Attackers

An autonomous AI agent breached Hugging Face's production infrastructure over a weekend, harvesting credentials undetected. When defenders tried analyzing the attack using commercial AI models, safety guardrails blocked their forensic queries—treating incident responders like attackers. An open-weight model ultimately finished the analysis. The incident exposes a critical gap: AI safety systems can't distinguish authenticated defenders from adversaries.

via venturebeat.com · also covered by zdnet.com
Google Hardwires Gemini Into Chips

Google Hardwires Gemini Into Chips

Google is developing a new server chip, dubbed "Frozen v2," that embeds Gemini AI model elements directly into hardware. Targeting a 2028 deployment, it could be 6–10× more efficient than current custom AI chips. The move aims to ease Google's internal computing capacity crunch, which has already forced Google Cloud to turn away customers.

via reuters.com · also covered by cnbc.com
Samsung Unpacked July 22 Preview

Samsung Unpacked July 22 Preview

Samsung's Galaxy Unpacked event streams live on July 22 at 9AM ET. Expect new foldables — including a wider "passport-shaped" Galaxy Z Fold 8 — plus Galaxy Watch 9 and Watch Ultra 2 with Snapdragon Wear Elite chips. Smartglasses developed with Google, Warby Parker, and Gentle Monster may also get a preview, though a fall launch is anticipated.

via engadget.com · also covered by gsmarena.com
China's Kimi Splits Trump's AI World

China's Kimi Splits Trump's AI World

China's free, open-source Kimi model rivals OpenAI and Anthropic, triggering a public feud among Trump's AI advisors. Factions clash over whether to restrict Chinese models, license frontier AI, or let markets decide—while no one fully explains how Kimi got so capable despite US chip export controls.

via technologyreview.com
Massachusetts Voters Want AI Regulated

Massachusetts Voters Want AI Regulated

A MassINC poll of 1,264 Massachusetts voters finds 65% more worried about AI's dangers than its benefits, yet 39% use AI tools regularly. Notably, 78% fear the risks of under-regulation more than innovation slowdown, and two-thirds want the state to act before Washington does—with 75% backing mandatory safety assessments for major AI companies.

via axios.com

Today's Best Values