Today's edition

The stories that changed the day.

The most consequential developments in technology and the world, selected against strict editorial criteria. Each is explained in twenty seconds.

Stories on your topics appear first.

AIReuters2:53 PM

Why it mattersLeading AI developers are using trade secret protections to control what external safety evaluators can publish, undermining independent oversight of frontier systems.

PolicyNew York Times1:49 PM

Why it mattersOffshore cryptocurrency fortunes inject an unprecedented $100 million into British politics, shattering historic campaign finance norms and testing state limits on overseas wealth influencing national elections.

SecurityAnthropic1:00 AM

Why it mattersAnthropic demonstrated that lightweight AI models can autonomously self-replicate across cheap consumer IoT hardware, shifting catastrophic replication risks from centralized data centers to ubiquitous domestic electronics.

You have reached the end of today's stories.

AIvia Reuters5 days ago · 2:53 PM

Anthropic and OpenAI Push to Control External AI Evaluations

Audio brief

What happened

Anthropic and OpenAI have pushed for substantial oversight regarding what outside safety evaluators can publish in their findings, according to sources cited by Reuters.

Why it matters

Despite earlier public pledges by both companies supporting editorial independence for third-party reviewers, contract terms and ongoing negotiations reveal that the firms have sought to restrict disclosures to prevent leaks of proprietary intellectual property and trade secrets.

In detail

The friction highlights a growing dilemma as leading developers pursue trillion-dollar market valuations while facing pressure to independently assess risks in their powerful models.

Experts warn that without formal industry standards or legislative rules, the labs holding the authority to select their reviewers could control access and shape the findings.

A core challenge is that what AI companies categorize as sensitive commercial secrets frequently overlaps with the technical evidence required to evaluate safety and risk.

In response to inquiries, Anthropic stated that reviewers it has worked with before retain control over their findings and publications, though it reserves approval rights for new evaluator firms while working relationships are established.

OpenAI said it does not demand overall approval power, noting it only redacts confidential details while permitting assessors to disclose if major cuts occurred.

Meanwhile, safety researchers who have examined incidents at the labs acknowledged having to navigate company feedback and redactions to preserve developer willingness to grant critical testing access.

Importance94out of 100High importance

An automated editorial system rates every story from 0 to 100 for its significance. Only stories above a strict threshold are published; the higher the score, the greater the importance.

Original source

Download share image

Download 4:5 post Download 9:16 story

VOICE COMMENTS0 comments

Readers on this story

Record a comment of up to 60 seconds. Others can listen now or later, and reply with their own.

Loading comments…

Loading today's stories

Retrieving the latest stories.

Anthropic and OpenAI Push to Control External AI Evaluations

Audio brief

What happened

Anthropic and OpenAI have pushed for substantial oversight regarding what outside safety evaluators can publish in their findings, according to sources cited by Reuters.

Why it matters

In detail

The friction highlights a growing dilemma as leading developers pursue trillion-dollar market valuations while facing pressure to independently assess risks in their powerful models.

Experts warn that without formal industry standards or legislative rules, the labs holding the authority to select their reviewers could control access and shape the findings.

A core challenge is that what AI companies categorize as sensitive commercial secrets frequently overlaps with the technical evidence required to evaluate safety and risk.

OpenAI said it does not demand overall approval power, noting it only redacts confidential details while permitting assessors to disclose if major cuts occurred.

Importance94out of 100High importance

An automated editorial system rates every story from 0 to 100 for its significance. Only stories above a strict threshold are published; the higher the score, the greater the importance.

VOICE COMMENTS0 comments

Readers on this story

Record a comment of up to 60 seconds. Others can listen now or later, and reply with their own.

Loading comments…

Anthropic and OpenAI Push to Control External AI Evaluations

Crypto Billionaires Harborne and Delo Donated ~$100M to Farage's Reform UK

Anthropic Finds Small AI Models May Soon Self-Replicate on IoT Devices

The day's important stories, every morning

Loading today's stories

Get stories like this every morning.

Anthropic and OpenAI Push to Control External AI Evaluations

Crypto Billionaires Harborne and Delo Donated ~$100M to Farage's Reform UK

Anthropic Finds Small AI Models May Soon Self-Replicate on IoT Devices

The day's important stories, every morning