We monitor AI safety policies and web content for substantive changes. Anonymous submissions: https://xitter.cf/t.co/5Ke9mIqOSM Run by @TheMidasProj
Joined May 2024
- Tweets188
- Following3
- Followers1.5K
- Likes117
The Midas Project Watchtower retweeted
If you saw MAGA influencers raging over a data center bill yesterday, you were probably looking at (another) paid astroturf.
We've documented this happening twice in two months, with nearly identical messaging and posts timed to hit just before key votes. Who is funding it? 🧵
The Midas Project Watchtower retweeted
1/ Did @OpenAI just break California’s new AI safety law?
The answer appears to be yes, and OpenAI could owe millions in fines. 🧵
The Midas Project Watchtower retweeted
Something strange happened on conservative Twitter on Thursday.
A dozen right-wing influencers suddenly became passionate about semiconductor export policy, posting nearly identical (and often false) attacks over a 27-hour period on a bill most people have never heard of.
🧵
The Midas Project Watchtower retweeted
I, too, made the mistake of *checks notes* taking OpenAI's charitable mission seriously and literally.
In return, got a knock at my door in Oklahoma with a demand for every text/email/document that, in the "broadest sense permitted," relates to OpenAI's governance and investors.
The Midas Project Watchtower retweeted
xAI’s ongoing quest to make all of its competitors look saintly-by-comparison is going well 😖
In the past few weeks, they’ve reached a new low: after releasing their safety policy three months late, they flagrantly violated that policy only one week later.
Details in 🧵
The Midas Project Watchtower retweeted
OpenAI once said its nonprofit would be entitled to "the vast majority" and "all but a fraction" of the wealth it generates.
Now, in their new restructuring, they are saying it will be entitled to only 20%. (~$100b out of a $500b valuation).
From "Nearly all" to "one fifth" 🙄
The Midas Project Watchtower retweeted
NEW: Anthropic will start training its AI models on user data, including new chat transcripts & coding sessions, unless users choose to opt out by 9/28 (it's a pop-up window that will give you the choice). It’s also extending its data retention to 5 years.
theverge.com/anthropic/76750…
The Midas Project Watchtower retweeted
🚨 Breaking: A group of 100+ Nobel laureates, professors, whistleblowers, public figures, artists, and nonprofit organizations just released a letter asking OpenAI to tell the truth about its restructuring.
Here’s what they had to say: 🧵
The Midas Project Watchtower retweeted
In 2024, @xai promised the governments of the UK and South Korea that it would release a frontier safety framework by February.
By February, xAI released an incomplete draft, and promised a new version "within three months."
Now, they've missed the deadline once again.
The Midas Project Watchtower retweeted
We're in @FastCompany today 🗞️ Check it out to learn more about how The Midas Project's Watchtower program (@SafetyChanges) is holding AI developers accountable by tracking corporate policy changes.
fastcompany.com/91304014/thi…
The Midas Project Watchtower retweeted
This isn't a change in substance and has caused some confusion - we're working on a fix. We continue to follow the White House Voluntary Commitments.
The Midas Project Watchtower retweeted
I asked our team about this! When we launched privacy.anthropic.com the content was moved into separate knowledge bases for Consumer & Commercial. Each paragraph remains, but in its respective knowledge base. (In your left img, Consumer is line 20 & Commercial is line 25) [...]
The Midas Project Watchtower retweeted
I'd let him correct the record himself but this is misleading: Todor asked for his name to be removed from the system card and we did.
The Midas Project Watchtower retweeted
OpenAI announced that its Safety & Security Committee, which the company introduced in May as it dealt with controversy over security processes, will become an independent board oversight committee. The group also just wrapped up a 90-day security review. cnbc.com/2024/09/16/openai-a…