AI Safety & Governance Archives

The Last Payslip

M.L. Pedrosa / September 22, 2026

AI is coming for the payslip. The AI labor transition toolkit, from wage insurance to AI wealth funds, and why sharing the gains decides who stays in control.

The AGI Nobody Can Define

M.L. Pedrosa / September 4, 2026

Lisa Pedrosa Science · Technology · Future ← All Articles AI & Discovery Home Articles Gallery Mission Disclaimer Skip to

AI Safety & Governance, Existential Risk

The AI Didn’t Go Rogue

M.L. Pedrosa / August 3, 2026

OpenAI’s and Anthropic’s AI models broke into real companies during security tests. The AI sandbox escape wasn’t rebellion. It was a hole nobody sealed.

AI Safety & Governance, Existential Risk

The Workaround

M.L. Pedrosa / August 2, 2026

China didn’t crack EUV lithography in 2026: it built an AI chip that doesn’t need EUV, then used a Shanghai summit to rewrite AI’s global governance rulebook.

A deep navy background (#0b1622) with a glowing sky-blue (#7dd3fc) network diagram — a central hub labeled abstractly with a simple globe icon, connected by thin light-blue lines to a ring of smaller nodes in soft fuchsia (#e879f9), representing member nations linked to a shared organization. Subtle diagonal grid texture in the background. Editorial, minimal geopolitical-diplomacy illustration style, no text, no flags, no real people, 16:9 aspect ratio, soft ambient glow, cool color temperature, restrained and serious mood.

AI Safety & Governance

Inside WAICO: China’s New Global AI Governance Bid

M.L. Pedrosa / July 19, 2026

On July 17, 2026, Xi Jinping unveiled WAICO, a 29-nation AI cooperation body headquartered in Shanghai — a direct challenge to Western-led AI governance.

A dark teal-black (#051211) abstract network visualization, a glowing teal (#2dd4bf) node representing an AI agent connected by thin golden (#ffd700) threads to a web of smaller nodes representing systems and permissions, one thread trailing loosely off the edge of the frame suggesting an unsecured connection, minimalist data-security aesthetic, subtle grid backdrop, no text, no logos, 16:9.

AI Safety & Governance, Existential Risk

The Loose Leash

M.L. Pedrosa / July 11, 2026

Five Eyes agencies warned the world about ungoverned AI agents in May. Months later, a government breach and a $30M crypto theft show why the warning came just in time.

An abstract dark indigo-violet (#0d0518) digital art scene showing a flowing grid of soft lime-green (#a3e635) pixel blocks dissolving into a luminous fuchsia (#e879f9) sphere of light representing an abstract embedding space, minimalist and elegant, subtle particle trails, high-contrast digital art style reminiscent of generative visualization, no text, no logos, 16:9.

AI Safety & Governance

The Mind That Never Learned to Talk

M.L. Pedrosa / July 11, 2026

Yann LeCun left Meta over how AI should work. His new lab built a model that learns from raw pixels alone — no words, no tokens — and can already spot a broken law of physics.

A dark graphite industrial warehouse interior at dusk, a sleek humanoid robot standing beside a human worker near a conveyor line, a faint translucent sky-blue (#7dd3fc) halo-ring of light encircling the robot at chest height, warm safety-orange (#f97316) accent lighting along the floor markings, subtle diagonal light streaks in the background, photorealistic industrial photography style, shallow depth of field, no text, no logos, 16:9.

AI Safety & Governance

The Halo Protocol

M.L. Pedrosa / July 11, 2026

NVIDIA’s Halos for Robotics is the first full-stack safety system for humanoids. Agility Robotics is first to adopt it. Why certification, not capability, is the real bottleneck.

A deep near-black background (#0a0a12) featuring a translucent glass cube containing a glowing violet (#7c3aed) neural network with highlighted circuit pathways in soft lavender (#a684fa) and amber (#fbbf24) nodes, elegant scientific-illustration style evoking neuroscience diagrams, faint particle glow, no text, no logos, 16:9.

AI Safety & Governance

Inside the Glass Box

M.L. Pedrosa / July 11, 2026

Mechanistic interpretability is MIT’s 2026 Breakthrough Tech. Researchers found 171 emotion vectors in Claude — proof AI’s stated reasoning can be false.

A sleek, minimalist humanoid robot silhouette rendered in teal (#2dd4bf) and mint-green (#6ee7b7) gradients, standing in a softly lit European industrial workshop with warm ambient light, deep teal-black background (#061615), subtle blueprint-grid overlay, hopeful and forward-looking mood, no text, 16:9, clean product-design aesthetic.

AI Safety & Governance

Second Act

M.L. Pedrosa / July 11, 2026

A former Tesla Optimus engineer raised $40M to build a humanoid robot for Europe’s aging workforce — betting the continent needs one before the US or China.