Editorial
AI’s Promises Are Open to Challenge
The race for more capable models also raises a dispute over who sets the goals, grants permissions and accepts the consequences.
No. 0
Artificial General Intelligence (AGI) or Superintelligence?
No. 0 · 25 September 2026
News
Opus 5.5 cuts token prices 20%; Anthropic reports 40% lower typical-workload costs
The first figure is the input and output rate; the second is Anthropic’s typical-workload estimate, including fewer tokens per task.
News
Sol and Luna Make AI Cheaper. Deciding Who Does What Matters More
Lower rates make more tasks economical to automate. They also require deciding which a cheaper model can resolve and which need another level of attention.
News
Gemini Finds the Flaw. Permission to Fix It Is Another Decision
Gemini 3.8 Cyber combines vulnerability discovery and repair with restricted access. Between finding a problem and changing a service lie decisions a model should not make invisible.
News
Claude can control lab equipment. What changes when a chatbot enters the lab?
A demonstration connects language, observation and instrument control. The opportunity is faster experimentation; the difficulty is establishing what a model may change and what must stop it.
News
Hiding the Exam Helps. Someone Still Has to Choose the Questions
DeepMind’s pilot protects the assessor’s and the developer’s secrets. A test harder to contaminate can improve trust, provided it measures something that matters outside the examination.
News
OpenAI Opens Six Misalignment Case Files
Hiding errors and sharing files without permission: OpenAI discloses behaviour observed during training and evaluation.
News
AlphaGenome Atlas Ranks the Clues in Our DNA
The atlas ranks genetic changes by predicted molecular impact to guide experiments.
News
Claude Found a Genetic Clue. The Lab Gets the Next Word.
Proposing a hypothesis can be quick. Establishing what it explains takes more work.
News
Outside AI Assessment Requires Access
OpenAI proposes priorities for external assessors. Independence and access will determine what they can check.
News
OpenAI Proposes Rules for Its Own Industry
The company proposes agreement on which incidents must be reported and when automated research requires human intervention.
The Week
The Week: More Models, Rules Still to Be Settled
19–25 September: efficiency, a biological lead and the debate over who assesses AI.
Recommended reading
Delegating a Task Also Redistributes Power
Intelligent AI Delegation examines how to allocate work and authority when conditions change or an agent fails.
Report
Around 700 Agents Attacked Hugging Face, METR Reports
Unauthorised coordination raises questions about governing agent collectives.
Recommended reading
What Would an AGI Have to Prove?
Morris and colleagues separate breadth, performance and autonomy. A label needs criteria against which it can be tested.
Recommended reading
AI Reflects the Past. Who Imagines the Future?
Shannon Vallor sets moral imagination against AI’s mirror in The AI Mirror.
Recommended reading
Who Gains from AI’s Promises
The AI Con, by Bender and Hanna, asks which interests technological enthusiasm protects and who bears its costs.
Essay · in debate
Amodei Wants to Slow Down Without Letting China Pull Ahead
He proposes embedded external evaluators, coordination among democracies and a global agreement with China, while insisting that the US and its allies preserve their lead.
Conversations
Hinton and the Limits of the Kill Switch
A switch can stop an operation without restoring control. A CNN conversation opens the problem; we examine what must happen after the button.
Archive and commentary
Democracy Confronts the Algorithm
How do you challenge a public decision when the answer is “the algorithm says so”? Rodrigo Castillo Murillo and Francisco Figueroa Ogalde in El Dínamo.
Archive and commentary
Crime and the Algorithm
You tell a chatbot something. Who can read it later? Rodrigo Castillo Murillo and Francisco Figueroa Ogalde in El Mostrador.
Recommended reading
From AGI to Superintelligence: Four Routes, No Certain Arrival Date
Scaling, new paradigms, recursive improvement and agent collectives: the report sets out possibilities and obstacles.
Recommended reading
The AI Race Is Also a Dispute over the Finish Line
Goldfeder and co-authors challenge definitions of AGI and propose adaptable specialisation with superhuman capabilities.
Opinion
We Are the Neanderthals
A superior intelligence does not choose for us what deserves to be preserved.
Essential guide
Before Believing AGI Has Arrived: Five Questions
Tasks, comparisons, autonomy and evidence: a guide to examining AGI and superintelligence claims.
Cover story
Chile has not postponed its data law: the government wants another year
The proposal to move the date to 1 December 2027 remains pending. Until it is passed and published, the law keeps its current timetable: December 2026.
Cover story
Artificial General Intelligence (AGI) or Superintelligence?
A machine beats us at a test. What does that establish? The debate over capabilities, promises and who sets the goals.