iiinova — where your digital self keeps evolving

The Midas Project @themidasproject.bsky.social · 05/08/2026

A full breakdown and diff of the changes are available here: themidasproject.com/watchtower/a... Anthropic does not provide a public archive of past FCF versions.

themidasproject.com

Anthropic | The Midas Project

Anthropic updated its Frontier Compliance Framework.

010

The Midas Project @themidasproject.bsky.social · 05/08/2026

Other changes include: - AI systems writing “large amounts of critical code” is also no longer specifically named as a factor that could activate loss of control risk tier 1 - Harmful manipulation is now included as an area subject to pre-launch risk analysis

themidasproject.com

Anthropic | The Midas Project

Anthropic updated its Frontier Compliance Framework.

110

The Midas Project @themidasproject.bsky.social · 05/08/2026

Most notably, the criteria for activating the automated AI R&D risk tier were updated to match Anthropic’s current Responsible Scaling Policy (v3.4).

100

The Midas Project @themidasproject.bsky.social · 05/08/2026

Company: Anthropic Date: July 24 Anthropic updated its Frontier Compliance Framework (FCF), the compliance-facing document that serves as Anthropic's framework under California's SB 53 and the EU General-Purpose AI Code of Practice.

120

The Midas Project @themidasproject.bsky.social · 29/07/2026

We just launched a YouTube channel. Our first video follows our investigation into AcutusWire, the fake news site using bots as reporters that was being funded by Leading the Future. Watch the full video here: youtu.be/2kcmP5leoLs

youtu.be

We Caught OpenAI’s Super PAC Funding a Bot-Run Fake Newsroom

YouTube video by Midas Project

041

Reposted by The Midas Project

Tyler Johnston @tylerjohnston.bsky.social · 03/06/2026

A few months ago, I found an anonymous sockpuppet account linked to the OpenAI/a16z super PAC. Now, @taylorlorenz.bsky.social and I have uncovered two more — and they're even more brazen than the first.

142

The Midas Project @themidasproject.bsky.social · 03/06/2026

New: Leading the Future, the pro-AI super PAC backed by OpenAI President Greg Brockman, appears to be linked to multiple sockpuppet accounts, including a false flag anti-AI activist Our latest for Model Republic with @taylorlorenz.bsky.social modelrepublic.substack.com/p/a-pro-ai-s...

modelrepublic.substack.com

A Pro-AI Super PAC's Secret Meme Sockpuppets

An OpenAI/a16z-affiliated super PAC appears to be linked to multiple sockpuppet accounts, attacking and parodying their opponents with lewd and offensive memes.

071

The Midas Project @themidasproject.bsky.social · 19/05/2026

Before SpaceX goes public, investors should see a full record of what has already gone wrong at xAI and a credible plan for managing the much bigger risks posed by the frontier AI capabilities the company may soon be releasing.

Highlighting SpaceX investors should therefore demand disclosure of the following.

110

The Midas Project @themidasproject.bsky.social · 19/05/2026

In April, deposed in his suit against OpenAI, Musk was asked about OpenAI's preparedness framework. "I don't know what a preparedness framework is," he answered. Which is awkward, because every major frontier AI lab publishes one of these. xAI has its own version.

100

The Midas Project @themidasproject.bsky.social · 19/05/2026

OpenAI's safety org has around 200 people. xAI's, according to the Washington Post, has two or three.

Shows XAI safety capacity and that safety is dead at the org XAI, according to the Washington Post in 2025. The safety team consisted of just "two or three people."

122

The Midas Project @themidasproject.bsky.social · 19/05/2026

xAI has not yet built Mythos-class models. Investors should be wondering whether it intends to, and if so, whether it has the organizational capacity to do so safely. On the second question, the evidence is not reassuring.

Comparison of XAI safety practices with those of Anthropic, OpenAI, and Google DeepMind.

210

The Midas Project @themidasproject.bsky.social · 19/05/2026

Mythos, Anthropic's most powerful and as-yet unreleased model, sits at the top of METR's chart, in a region where existing measurement tools are no longer reliable.

Frontier model time horizon on long software-engineering tasks (50% reliability). Source: METR.

100

The Midas Project @themidasproject.bsky.social · 19/05/2026

But that financial exposure only reflects xAI at its current capability level. The bigger issue is what will happen next.

100

The Midas Project @themidasproject.bsky.social · 19/05/2026

You probably know the broad strokes of what Grok has been up to. What jumps out when you look at the full record is the scale of the regulatory and legal response. It has grown large enough to represent real exposure for whoever buys into SpaceX.

Potential consequences of XAI safety practices.

110

The Midas Project @themidasproject.bsky.social · 19/05/2026

SpaceX is about to go public, and inside it now is xAI, a frontier AI lab that ranks behind its peers in every major published assessment of AI safety practices. What does that mean for investors? Our new report with Guidelight AI Standards, Encode, and LASST is out today.

Image of report: xAI: The Unpriced Risk in SpaceX's IPO.

131

The Midas Project @themidasproject.bsky.social · 12/05/2026

From @tylerjohnston.bsky.social: "... the more interesting question we can ask about Anthropic’s AI-generated training data disclosure is not “was it AI-generated?” but rather “is it slop?” substack.com/home/post/p-...

substack.com

Why is Anthropic’s training data disclosure AI-generated?

Anthropic and its peers don’t seem to be taking California’s data transparency law seriously.

011

The Midas Project @themidasproject.bsky.social · 24/04/2026

@tylerjohnston.bsky.social for Model Republic: www.modelrepublic.org/articles/the...

031

Reposted by The Midas Project

Tyler Johnston @tylerjohnston.bsky.social · 07/04/2026

I wasn't too shocked when an anon reply guy started pestering me + other orgs/journalists on X. Normally I’d ignore it, but I looked closer when they began running paid ads. I did not expect to trace the account back to OpenAI’s political machine! My latest for @themidasproject.bsky.social

A grid of tweets from @DoomerDaylight on twitter

165

The Midas Project @themidasproject.bsky.social · 07/04/2026

An anonymous account spent months attacking OpenAI’s critics with misleading claims and paid ads. We found links to Targeted Victory, the firm at the center of OpenAI's $125 million political operation. www.modelrepublic.org/articles/is-...

modelrepublic.org

Is this OpenAI’s anonymous Twitter sockpuppet account? - Model Republic

062

The Midas Project @themidasproject.bsky.social · 26/03/2026

Read our full article on why voluntary frameworks are failing and how independent standards and auditing could improve the situation: www.modelrepublic.org/articles/ai-...

011

The Midas Project @themidasproject.bsky.social · 26/03/2026

A regulatory regime built around self-authored rules and inadequate compliance monitoring will naturally converge on weak, vague, and ineffective standards. We need something stronger.

100

The Midas Project @themidasproject.bsky.social · 26/03/2026

We are seeing this play out across the industry. Findings from @safetychanges show that OpenAI, Google, xAI, and Anthropic have all recently evaded, ignored, or weakened their safety rules to keep pace.

110

The Midas Project @themidasproject.bsky.social · 26/03/2026

Between California’s SB53, New York’s RAISE Act, and the EU AI Act, multiple regulatory regimes have emerged that are built around companies writing their own rules. But as commercial pressures are increasing, this system is failing.

100

The Midas Project @themidasproject.bsky.social · 26/03/2026

When California passed a law requiring AI companies to follow their safety frameworks, Anthropic had a choice to make. Instead of making their strict, voluntary rules legally binding (as they had indicated they wanted), the company wrote a second, weaker rulebook in its place.

141

The Midas Project @themidasproject.bsky.social · 19/03/2026

So long as this race continues, we should expect continued weakening of standards + companies taking every opportunity they can to minimize liability. We'll keep tracking how their commitments evolve. Follow @safetychanges.bsky.social for document-level analysis.

000

The Midas Project @themidasproject.bsky.social · 19/03/2026

When updating the RSP to its largely-weaker v3.0 last month, Jared Kaplan told TIME “We didn't really feel, with the rapid advance of AI, that it made sense for us to make unilateral commitments … if competitors are blazing ahead.”

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

If the most safety-conscious lab treats binding legislation as an exercise in minimizing liability rather than an opportunity to codify best practices, the “race to the top” they were hoping for collapses.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

This sets a grim precedent. So far, no other companies have created a separate (weaker) policy for compliance purposes, but seeing Anthropic’s example, they might soon.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

The result is a dual-track system with a voluntary RSP containing detailed commitments yet imposing no liability, alongside a vague and discretionary policy that absorbs all the liability.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

The new FCF also strips out key governance mechanisms found in the RSP. It removes the detailed Capability and Safeguard Reports and eliminates the formal oversight role previously held by the Long-Term Benefit Trust regarding model deployment decisions.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

This if-then structure was the core strength of the original RSP, and now it is gone (and not only from the FCF, but also from the RSP itself insofar as they have revoked their commitment to always follow the voluntary RSP). time.com/7380854/excl...

time.com

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

The FCF is like the RSP if you tore out much of its substance. It drops the RSP's if-then structure by attaching no specific, binding mitigations to its risk tiers. The text simply states that mitigations "may be determined when the relevant risk tier is reached."

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

They released a second (and largely weaker + more vague) document that would substitute for the RSP for compliance purposes, the Frontier Compliance Framework (FCF). By doing so, they ensured that all the promises within the RSP weren’t rendered enforceable by SB 53.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

You’d think they’d be excited to have succeeded in codifying the RSP model. But, just before SB 53 took effect, Anthropic significantly watered down its obligations with a last-minute change.

110

The Midas Project @themidasproject.bsky.social · 19/03/2026

Anthropic endorsed the bill, suggesting it would hold all companies to the standard of Anthropic’s RSP. In the past, they have discussed the RSP as a prototype for regulation (even as the substance of their policy has changed).

110

The Midas Project @themidasproject.bsky.social · 19/03/2026

Over time, other companies adopted similar policies, and by 2025, this standard was put into law: California’s newly passed SB 53 *required* frontier AI developers to publish and follow safety frameworks like RSPs.

100

The Midas Project @themidasproject.bsky.social · 19/03/2026

In 2023, Anthropic worked with METR to adopt the first Responsible Scaling Policy (RSP). The RSP was a set of “if-then” commitments (if a model hits capability X, implement safeguard Y, or pause). It was the most rigorous commitment any frontier AI lab had made at the time.

110

The Midas Project @themidasproject.bsky.social · 19/03/2026

Over at @safetychanges.bsky.social, we just released an analysis of Anthropic’s recent, quiet change to their Frontier Compliance Framework. But a more interesting story (which many missed) is the fact that this policy exists at all — and how it minimizes liability for the company. 🧵

110

The Midas Project @themidasproject.bsky.social · 26/02/2026

Musk was one of the loudest voices warning about AI and the risks of autonomous warfare. Now he’s set the precedent every other AI company is being pressured to follow.

050

The Midas Project @themidasproject.bsky.social · 26/02/2026

Anthropic has refused, blocking surveillance and autonomous weapons use. Secretary of Defense Hegseth threatened to invoke the Defense Production Act to force them to comply and gave them until 5:01pm Friday to back down.

153

The Midas Project @themidasproject.bsky.social · 26/02/2026

Musk didn’t just go back on what he promised, he also opened the door for the Pentagon to pressure every other AI company to accept the same terms. Google and OpenAI are reportedly in talks with the Pentagon to expand their model access for unrestricted DoD use.

121

The Midas Project @themidasproject.bsky.social · 26/02/2026

Since then, Musk has aided the government surveillance he once challenged: X continues sharing data with law enforcement through Dataminr.

132

The Midas Project @themidasproject.bsky.social · 26/02/2026

It’s not just weapons. In 2023, X petitioned SCOTUS to be able to disclose how often the government demanded its users' data, arguing that "surveillance of electronic communications is both a fertile ground for government abuse and a lightning-rod political topic…"

100

The Midas Project @themidasproject.bsky.social · 26/02/2026

But in 2026, SpaceX and xAI entered the Pentagon's competition to build autonomous drone swarms — exactly what he pledged not to do. A Pentagon official confirmed that the drones will be used for offensive purposes.

101

The Midas Project @themidasproject.bsky.social · 26/02/2026

In 2018, Musk signed a pledge alongside DeepMind's founders, 3,800+ individuals and 274 organizations: "We will neither participate in nor support the development, manufacture, trade, or use of lethal autonomous weapons."

221

The Midas Project @themidasproject.bsky.social · 26/02/2026

This week, xAI agreed to let the Pentagon use Grok in classified systems under the “all lawful use” standard, enabling Grok to be used for autonomous weapons and domestic mass surveillance (something other companies have objected to).

132

The Midas Project @themidasproject.bsky.social · 26/02/2026

Elon Musk spent a decade warning the world about killer robots and AI surveillance, including in a 2017 open letter: "We do not have long to act. Once this Pandora's box is opened, it will be hard to close." Now he’s the one opening it. 🧵

142

The Midas Project @themidasproject.bsky.social · 25/02/2026

That’s why we’re keeping track of the AI industry’s influence operations in 2026, both disclosed and discreet. Read our full story: www.modelrepublic.org/articles/ano...

010

The Midas Project @themidasproject.bsky.social · 25/02/2026

AI policy is too important for the public conversation to be shaped by discreet influence campaigns and heavy spending from an industry that has billions of dollars riding on the regulatory environment remaining friendly to its interests.

130