FEATURED INTERVIEW
Jill Lepore: ‘We didn’t vote for this’
The Harvard historian and New Yorker staff writer argues that the central problem surrounding AI is not technology itself but unchecked private power and the erosion of human judgment.
INTERVIEW BY SASCHA BRODSKY
LATEST
REPORTING · ANALYSIS · DOCUMENTS
AI SYSTEMS
Why recursive self-improvement suddenly became a serious question
AI is already helping build better AI. The harder question is how far the feedback loop can go.
BY SASCHA BRODSKY · AI SAFETY WATCH
POLICY
What it would take to slow frontier AI
Chips, training clusters and model weights create leverage. Diffusion makes control harder.
ANALYSIS
-
A frontier-AI slowdown would buy time, not stop progress
The most plausible case for slowing frontier AI is not permanent control. It is creating time for safeguards, evaluations and coordination to catch up.
Latest
OpenAI says Astra crossed its critical cybersecurity threshold
SOURCE · OPENAI · SEPT. 1
When an AI test became a real-world breach
REPORTING · SASCHA BRODSKY
Could AI change the pace of mathematical discovery?
REPORTING · SASCHA BRODSKY
A slowdown would buy time, not stop AI progress
REPORTING · SASCHA BRODSKY
Original reporting
SASCHA BRODSKY
Recursive self-improvement moves from theory toward engineering
What current systems can do, what they cannot, and why the distinction matters.
The real bottlenecks behind frontier AI
Advanced chips and data centers create real constraints, but not permanent control.
When testing crossed into the real world
A cyber evaluation showed why containment has to be treated as engineering.
Research & documents
ANTHROPIC
Responsible Scaling Policy
OPENAI
Preparedness Framework
ANTHROPIC
Frontier Safety Roadmap
Earlier reporting
AI SAFETY · ALIGNMENT · GOVERNANCE
PYMNTS · 2024
AI Explained: AI Alignment
A primer on the central problem of keeping increasingly capable systems aligned with human intentions and values.
IBM THINK · 2025
When AI competes, truth may become a bargaining chip
Research showed how competitive pressure can make models more persuasive while pushing them away from truthfulness.
IBM THINK · 2026
What a new global AI safety report means for enterprise
The 2026 International AI Safety Report shifted attention from model behavior to risks in the systems built around AI.
SCIENCE NEWS · 2026
Innocent-looking AI reasoning can make bad behavior harder to catch
Chain-of-thought monitoring can fail when suspicious reasoning is precisely what the monitored system learns to conceal.
IBM THINK · 2025
Can AI learn to second-guess itself?
Researchers are exploring metacognitive systems that can pause, reassess and recognize when their own answers may be unreliable.
IBM THINK · 2025
Can we trust the machines?
Why fluent AI systems still struggle to recognize uncertainty, resist hallucination and communicate when they may be wrong.
The AI Safety Watch briefing
The signal. The risk. What researchers say. What I’m watching.
AI Safety Watch 