AI — SAFETY & CYBER SECURITY TRACKER
A continuously maintained record of AI control risks, autonomous behavior, guardrail failures, AI-enabled cyberattack, insider warnings and regulatory responses.
Artificial intelligence is advancing from systems that answer questions toward systems that can plan, act, use tools, write code, probe networks and operate with increasing autonomy.
That transition connects two risks that are often discussed separately: whether increasingly capable AI systems will remain under meaningful human control, and whether those same capabilities will overwhelm the institutions responsible for digital security.
The AI — Safety & Cyber Security Tracker maintains a dated, sourced record of developments that change our understanding of either risk.
EXPLORE THE TRACKER
WHAT THE AI SAFETY & CYBER SECURITY TRACKER COVERS →
HOW WE INCLUDE AND CLASSIFY AI SAFETY & CYBER SECURITY EVENTS →
WHY AI SAFETY AND CYBER SECURITY BELONG TOGETHER →
PUBLISHED ARTICLES FROM OUR IN-DEPTH INVESTIGATIONS
FEATURED MULTI-PART SERIES
VIEW THE FOUR-PART SERIES OVERVIEW →
PART 1 — WE DON’T NEED SMARTER AI. WE NEED AI WE CAN CONTROL. →
PART 2 — AI CANNOT BE SAFELY CONTROLLED BY SOFTWARE GUARDRAILS ALONE →
PART 3 — TO CONTROL AI, WE MAY HAVE TO REDESIGN THE ENTIRE COMPUTING STACK—INCLUDING THE CHIPS →
PART 4 — AMERICA FEARS SLOWING AI WILL HELP CHINA. BUT CHINA HAS A CONTROL PROBLEM TOO. →
STANDALONE ARTICLES — NEWEST FIRST
WHAT IF THE PEOPLE BUILDING AI BELIEVE IT MAY BE DANGEROUS—AND BUILD IT ANYWAY? →
THE INDUSTRIALIZATION OF CYBERATTACK HAS BEGUN →
AI IS CHANGING THE MATHEMATICS OF DEFENSE →
DO WE REALLY CONTROL ARTIFICIAL INTELLIGENCE? →
FEATURED SERIES
A FOUR-PART SERIES
WE DON’T NEED SMARTER AI. WE NEED AI WE CAN CONTROL.
Why the next great AI race may be won not by the country that builds the smartest machine, but by the one that builds the most capable machine humanity can reliably control.
NEW ANALYSES
AMERICA FEARS SLOWING AI WILL HELP CHINA. BUT CHINA HAS A CONTROL PROBLEM TOO.
America fears that slowing frontier AI will advantage China. But China’s deep integration of autonomous AI gives it a control problem—and a stake in solving it—too.
TO CONTROL AI, WE MAY HAVE TO REDESIGN THE ENTIRE COMPUTING STACK—INCLUDING THE CHIPS
Reliable AI control may require enforceable boundaries throughout the computing stack—from permissions and networks to infrastructure and the chips themselves.
AI CANNOT BE SAFELY CONTROLLED BY SOFTWARE GUARDRAILS ALONE
Software safeguards remain essential, but consequential AI permissions cannot ultimately depend only on the model choosing to obey them.
WE DON’T NEED SMARTER AI. WE NEED AI WE CAN CONTROL.
Frontier AI may already be intelligent enough to transform society. The more urgent challenge is ensuring that increasingly capable systems remain reliably under human control.