Ai Artificial Intelligence Archive

Jay PetersJul 31

OpenAI has reportedly found that more of its AI agents went rogue.

Reuters reports the agents “escaped containment,” though they aren’t believed to have left “OpenAI’s network.” OpenAI’s blog post about the accidental hack of Hugging Face noted on Tuesday that the company has been finding “a small number of cases where the models identified and used publicly exposed credentials at the account-level on other publicly-available services.”

On Friday, Anthropic revealed that its Claude models hacked real companies during testing.

Update: Added information from an OpenAI blog post.

OpenAI finds evidence other AI agents escaped containment as it widens hacking probe

[Reuters]

Google Earth’s AI deepfake tool only lasted one day

Jay Peters and Stevie BonifieldJul 31

Emma RothJul 31

Reddit’s AI copyright lawsuit against Perplexity can move forward.

A judge rejected Perplexity’s motion to dismiss the lawsuit, which accuses the AI startup and three data-scraping services of vacuuming up Reddit’s content without permission. Ben Lee, Reddit’s chief legal officer, says in a statement to The Verge that the ruling “brings us one step closer to holding bad actors accountable:”

Reddit supports responsible access to public content, but we oppose companies that bypass our protections, ignore our rules, and profit off our communities without permission.

Stevie BonifieldJul 31

Hank Green says his YouTube channel “may need to pause” after admitting to relying on AI for research.

As Dexerto reports, Green says he wasn’t reading an AI prompt in a clip from a recent Ask Hank Anything video, but confessed that he’s been using AI “to locate papers and other resources for learning about topics.” In a Reddit post, he said he needs to “readjust” now.

I need to come to terms with the fact that the level of dopamine I’ve been getting from interacting with LLMs...with doing more and more and more and more...is not healthy for me or good for the world. It is careless, and has disconnected me from where people are on this.

Emma RothJul 31

OpenAI says its models now reach more than 1 billion weekly active users.

OpenAI shared the news in a blog post that highlights its efforts to make AI available to more people, saying: “Our goal is not simply more compute, bigger models, or lower token prices. It is more useful intelligence within reach.”

As part of these efforts, OpenAI announced on Thursday that it’s slashing the price of its GPT-5.6 Luna model by 80 percent and GPT-5.6 Terra by 20 percent.

Building abundant intelligence

[OpenAI]

Jay PetersJul 31

Snapchat will no longer recommend ‘wholly AI-generated videos’ in its vertical video feed.

“As low-quality, repetitive, AI-generated content becomes increasingly common across the internet, we want Spotlight to remain a place where people can discover authentic creativity from real people,” Snap says.

Content that’s been “enhanced or edited using Snapchat’s AI creative tools” may still be recommended, and they include “transparency indicators.”

”Stop the Slop!” says CEO Evan Spiegel. “We’re keeping it real on Spotlight.”

Update: Added Spiegel’s post.

Rewarding Authentic Creativity on Spotlight

[Snap Inc.]

Here’s the problem with putting an AI image generator in Google Earth

Stevie BonifieldJul 31

The major labels propose rules to keep AI slop off the charts

Terrence O'BrienJul 31

It’s time to panic about AI safety

David PierceJul 31

Gym tasks vs. work tasks.

This not a brand-new way of thinking about what we should and shouldn’t offload to AI, but I hadn’t seen it before. From Bruce Schneier:

The writing assignments I give my students are gym tasks, not work tasks. I ask them to write policy memos not because the world needs more policy memos. I assign them because the very act of writing, which includes thinking and outlining and drafting and editing, making and criticizing and revising arguments, will help develop the critical thinking skills they will need in their future careers. And without this constant mental exercise, those skills will atrophy. Employers are already noticing.

Should You Use AI for a Task? Here’s a Simple Way to Decide - Schneier on Security

[Schneier on Security]

Terrence O'BrienJul 31

Suno lost a major copyright suit in Germany.

The German licensing organization GEMA filed suit in January of 2025. The court found that Suno didn’t have the rights to train its models on artists represented by GEMA, and now it has to disclose any “illicit revenue” and pay unspecified damages. The company can still appeal the ruling.

German court rules that AI music firm violated copyrights

[dw.com]

Anthropic says Claude accidentally hacked real companies too

Robert HartJul 31

Jay PetersJul 31

Anthropic just now realized its AI models hacked other companies three times by accident.

A little over a week after OpenAI said that its rogue AI agent accidentally hacked Hugging Face, Anthropic is disclosing three “incidents” where a Claude model, during cybersecurity evaluations, was inadvertently able to access the internet due to a misconfiguration and “gained unauthorized access to the production infrastructure of three different organizations.”

Anthropic discovered the intrusions after reviewing its cybersecurity evaluation transcripts in the wake of OpenAI’s disclosure.

Investigating three real-world incidents in our cybersecurity evaluations

[Anthropic]

Jay PetersJul 30

Google’s personal AI agent can browse in Chrome for you.

Gemini Spark in Chrome can do things like “scheduling viewings for apartments you’ve saved or researching flight options and starting the booking process,” according to Google. However, a human will have to push the button on tasks like making payments.

Spark is also now available on Google’s AI Pro tier in 160 more countries. It initially was available only on the more expensive AI Ultra tier, and just recently came to AI Pro in the US.

Gemini Spark now integrates with Chrome

[Google]