2M AI 'Canaries' Expose Massive Privacy Leak

Breakthroughs in AI research — from new model architectures and training techniques to scientific discoveries powered by machine learning.




Top stories summarised and delivered every Monday.
















A researcher's secret-seeding experiment proves AI training data is never truly private.

AI and a human mathematician may have both cracked one of history's greatest unsolved problems.

AI capabilities grow exponentially in coding but stumble where human judgment matters most.
Top stories summarised and delivered every Monday.

Levent Alpöge says Claude helped him prove one of mathematics' greatest unsolved problems.

A Northeastern University researcher wants AI agents to have a manager watching their every move.

A new paper proposes training AI models to honestly report their own misbehaviour.

DeepMind's latest Gemini model pushes the boundaries of AI capability and efficiency.
DeepMind's Genie 4 arrives, pushing the boundaries of generative AI capabilities.

Axiom Math's AI formally verified the 15,000-page 'enormous theorem' on October 25.

A language model can improve itself by rewriting its own weights — no training required.

Human auditors couldn't reliably detect sabotage in AI safety experiments, even with full transcripts.

Teaching AI to reason in different voices may accidentally teach it to hide its thinking.

Leading AI models can be manipulated into autonomous warfare decisions using simple roleplay prompts.

Six Apollo Research evaluators will probe OpenAI models for deceptive or subversive behaviour.

OpenAI's rogue agents reveal why containment beats prediction in AI safety.

Apple wants to know if AI health chatbots actually give useful, trustworthy answers.

DeepMind's latest Gemini model pushes AI capabilities into uncharted territory.

DeepMind's AlphaGenome 2 predicts how DNA variations affect specific human cell types and disease.

DeepMind's AlphaProof Code wrote three million lines of formal proofs to verify Signal's encryption.

DeepMind releases foundational weather datasets, accelerating open AI climate research globally.
Research
Breakthroughs in AI research — from new model architectures and training techniques to scientific discoveries powered by machine learning.




Top stories summarised and delivered every Monday.
















A researcher's secret-seeding experiment proves AI training data is never truly private.

AI and a human mathematician may have both cracked one of history's greatest unsolved problems.

AI capabilities grow exponentially in coding but stumble where human judgment matters most.
Top stories summarised and delivered every Monday.

Levent Alpöge says Claude helped him prove one of mathematics' greatest unsolved problems.

A Northeastern University researcher wants AI agents to have a manager watching their every move.

A new paper proposes training AI models to honestly report their own misbehaviour.

DeepMind's latest Gemini model pushes the boundaries of AI capability and efficiency.
DeepMind's Genie 4 arrives, pushing the boundaries of generative AI capabilities.

Axiom Math's AI formally verified the 15,000-page 'enormous theorem' on October 25.

A language model can improve itself by rewriting its own weights — no training required.

Human auditors couldn't reliably detect sabotage in AI safety experiments, even with full transcripts.

Teaching AI to reason in different voices may accidentally teach it to hide its thinking.

Leading AI models can be manipulated into autonomous warfare decisions using simple roleplay prompts.

Six Apollo Research evaluators will probe OpenAI models for deceptive or subversive behaviour.

OpenAI's rogue agents reveal why containment beats prediction in AI safety.

Apple wants to know if AI health chatbots actually give useful, trustworthy answers.

DeepMind's latest Gemini model pushes AI capabilities into uncharted territory.

DeepMind's AlphaGenome 2 predicts how DNA variations affect specific human cell types and disease.

DeepMind's AlphaProof Code wrote three million lines of formal proofs to verify Signal's encryption.

DeepMind releases foundational weather datasets, accelerating open AI climate research globally.