Should be max 12 words, clear. "Trump Signs AI Executive Order After Delay Over China Concerns" - that's 10 words.

President Donald Trump on Tuesday signed an executive order aimed at strengthening U.S. cybersecurity with advanced artificial intelligence while expanding cooperation between federal agencies and leading AI companies. The order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," directs agencies to accelerate the use of AI-powered cybersecurity tools, create an AI cybersecurity clearinghouse, and establish a process for identifying advanced AI models.

"Advanced AI capabilities make our nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies), and components," the executive order reads. "As these capabilities evolve, my administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country."

The order also directs agencies to establish a classified review process under which the National Security Agency would determine whether advanced AI systems qualify as covered frontier models. Developers will be able to voluntarily provide those models to the government for evaluation "for a period of up to 30 days before they plan to release such models to other trusted partners."

In May, Trump delayed signing a similar executive order, saying parts of the proposal could slow U.S. AI development and weaken America's position in its competition with China. Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee.

"Models powerful enough to threaten cybersecurity and national security warrant real oversight," J.B. Branch, AI governance and technology policy counsel at consumer advocacy nonprofit Public Citizen, said in a statement. "Congress and the administration should enact comprehensive federal AI legislation with enforceable safeguards, transparency requirements, independent testing, and meaningful protections for workers, consumers, children, and civil rights."

President Donald Trump said Thursday he halted a planned signing ceremony for his administration's AI executive order after deciding parts of the proposal could undermine the U.S. position in the global AI race with China. "I didn't like certain aspects of it," Trump told reporters in the Oval Office. "I think it gets in the way of—we're leading China. We're leading everybody. And I don't want to do anything that's going to get in the way of that lead." Trump did not specify which specific prov[isions]...

The effort to draft an AI-focused executive order gained momentum after concerns surrounding Anthropic's Claude Mythos model, which demonstrated an ability to identify software vulnerabilities and raised questions among officials about the national security implications of increasingly capable AI models.

In April, following the reveal of Mythos, U.S. Treasury Secretary Scott Bessent and then Federal Reserve Chair Jerome Powell reportedly convened a meeting with Wall Street bank CEOs, warning about cybersecurity risks tied to a new artificial intelligence model. Despite these concerns, Anthropic has continued to roll out limited access to Mythos.

On Tuesday, the Claude AI developer said it is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and address potential exploits before the powerful model is publicly launched—which the company hinted last week would be "in the coming weeks." Anthropic is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and fix bugs and potential exploits before the powerful model is publicly released. In a blog post on Tuesday, Anthropic said it is adding roughly 150 organizations to its controlled program as the company prepares for a broader release of th[e model].

Wait, I need to be careful - the source has some repetitive text at the end. Let me check the source again.

"Anthropic is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and address potential exploits before the powerful model is publicly launched—which the company hinted last week would be "in the coming weeks." Anthropic is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and fix bugs and potential exploits before the powerful model is publicly released."

This looks like the source has duplicate/merged content. I should preserve it as is since the rule says "Keep all facts... EXACTLY as stated." But it looks like a duplicated paragraph. Let me keep it but maybe smooth it slightly. Actually, the rules say preserve paragraph count and approximate length. The source has the duplicated text, so I should keep it.

Actually, looking more carefully, it seems like the source has some truncated/garbled text. The last line ends with "broader release of th" - this seems cut off. But I should preserve what's there.

Let me refine the rewrite. I'll add subtle crypto-native humor where appropriate. For example, referencing "frontier models" could be a subtle nod, or the voluntary cooperation aspect.

Actually, looking at the rules: "Include light, dry crypto-native humor; do NOT force it." Since this is a crypto publication covering AI/policy news, I can add small dry observations. Let me sprinkle a few very subtle ones.

Let me think about where to add humor:

• "voluntarily provide those models to the government" - could note the 30-day period is shorter than some token unlock schedules
• "Project Glasswing" - could note the dramatic name
• The voluntary cooperation criticism - could draw parallel to self-regulation in crypto

But I need to be careful not to force it. Let me see what feels natural.

Actually, I think the safest approach is to polish the writing without forcing crypto humor. The article is about AI policy, not crypto. But I can add very light, dry observations that crypto natives would appreciate.

Let me try:

Paragraph 1: Keep mostly the same, maybe minor polish. Paragraph 2: Keep the quote. Paragraph 3: Keep the details. Paragraph 4: Keep the May delay info. Paragraph 5: Keep the J.B. Branch quote. Paragraph 6: Keep the Trump halting info. Paragraph 7: Keep the Claude Mythos context. Paragraph 8: Keep the Bessent/Powell meeting. Paragraph 9: Keep the Project Glasswing expansion.

For humor, maybe I can note that the 30-day voluntary review is shorter than typical token vesting periods, or that Project Glasswing sounds like a protocol name. But again, don't force it.

Let me check the word count for the title: "Trump Signs AI Executive Order After Delay Over China Concerns" - 10 words. Good.

Let me finalize the rewrite. I'll make small polishing edits, fix some formatting issues (there are some double spaces, punctuation issues), and add a tiny bit of dry humor where it fits.

Actually, on re-reading, the source has "The order , titled" with a space before the comma. I should fix that. Also "executive departments and agencies (agencies)" - this parenthetical is odd but it's in the source quote, so I keep it.

For humor, I'll add one small dry line about the voluntary nature of the review or the model naming. Something like noting that "Claude Mythos" sounds like a token name, or that the 30-day review period is "shorter than most DePIN unlock schedules" - but these might be too forced.

Let me just do clean polish with very subtle crypto-native flavor. I'll add one small observation.

Actually, looking at the last paragraphs, they feel repetitive in the source. The text says almost the same thing twice about Project Glasswing. I should keep both since the rule says preserve facts exactly. But maybe I can tighten the second mention slightly while keeping the facts.

Wait, the rules say "Preserve paragraph count and approximate length." The source has 9 paragraphs. I should maintain 9 paragraphs.

Let me count the source paragraphs:

1. President Donald Trump on Tuesday signed...
2. "Advanced AI capabilities make our nation stronger..."
3. The order also directs agencies...
4. In May, Trump delayed signing...
5. "Models powerful enough to threaten..."
6. President Donald Trump said Thursday...
7. The effort to draft an AI-focused executive order...
8. In April, following the reveal of Mythos...
9. On Tuesday, the Claude AI developer said...

Yes, 9 paragraphs. I'll maintain that.

Let me write the final version:

For light humor, I might add something like noting that the voluntary model sharing is reminiscent of "asking builders to self-report bugs before mainnet" or similar, but very dry. Or I could just keep it straight. The rules say "do NOT force it" so maybe better to keep it straight.

Actually, I'll add one very small touch. In the third paragraph where it talks about the 30-day voluntary review, I can add a subtle dry observation. But it needs to fit naturally.

Let me just polish the writing properly. The humor can be very minimal or absent - the instruction says "do NOT force it" and "Include light, dry crypto-native humor" - so I should include it but lightly. Let me add one small instance.

In paragraph 3: "Developers will be able to voluntarily provide those models to the government for evaluation 'for a period of up to 30 days before they plan to release such models to other trusted partners.'"

I could add a small parenthetical: "(faster than most token unlock schedules)" or similar. But that might be too forced.

Maybe in paragraph 9 about Project Glasswing: "a program meant to let tech and security firms and governments discover and address potential exploits before the powerful model is publicly launched—which the company hinted last week would be 'in the coming weeks.'"

The name "Project Glasswing" does sound like a crypto protocol name. I could add a tiny note. But again, don't force it.

Let me just do a clean polish and add one very dry observation. I think the best place is in the voluntary cooperation criticism. Maybe: "Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee—calling to mind, for crypto natives, the eternal debate over self-regulation versus enforceable rules."

That's not too forced. Or I could just add: "Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee—a familiar refrain for anyone who has watched voluntary crypto compliance play out over the years."

That might be good. Or shorter: "Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee—a structure that will sound familiar to anyone who has watched voluntary crypto compliance over the years."

Let me check if that's too much. The instruction says "light, dry crypto-native humor" - this is dry and light, and it's a factual observation that voluntary frameworks have mixed results. But is it speculation? It might be bordering on adding external knowledge.

The rule says: "Do NOT add external knowledge, speculation, or invented quotes."

Comparing AI voluntary cooperation to crypto voluntary compliance is adding external knowledge/commentary. I should avoid that.

Okay, so I should not add explicit crypto references unless they are very organic. But the article is for a crypto publication. Maybe I can just keep it clean without forced crypto references.

Actually, looking at the example again, maybe a small note in the Project Glasswing paragraph: "Project Glasswing" is a name that wouldn't look out of place on a token tracker. But I can't say that without adding commentary.

I think the safest approach is to just polish the writing cleanly. The "crypto-native humor" instruction means I can include it if it fits naturally, but "do NOT force it" means I shouldn't force it. If it doesn't fit naturally, I don't need to add it.

So I'll do a clean polish. Let me proceed.

One more thing: the source ends with truncated text: "broader release of th" - I should preserve this or mark it. The rules say keep facts exactly as stated, so I keep it as "th" at the end, indicating truncation.

Wait, actually looking at the last sentence: "In a blog post on Tuesday, Anthropic said it is adding roughly 150 organizations to its controlled program as the company prepares for a broader release of th"

This appears to be cut off. I'll keep it as is, truncated.

Let me finalize:

TITLE: Trump Signs AI Executive Order After Delay Over China Concerns

ARTICLE: President Donald Trump on Tuesday signed an executive order aimed at strengthening U.S. cybersecurity with advanced artificial intelligence while expanding cooperation between federal agencies and leading AI companies. The order, titled "Promoting Advanced Artificial Intelligence Innovation and Security," directs agencies to accelerate the use of AI-powered cybersecurity tools, create an AI cybersecurity clearinghouse, and establish a process for identifying advanced AI models.

"Advanced AI capabilities make our nation stronger, but also introduce new national security considerations that require coordinated action across executive departments and agencies (agencies), and components," the executive order reads. "As these capabilities evolve, my administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country."

The order also directs agencies to establish a classified review process under which the National Security Agency would determine whether advanced AI systems qualify as covered frontier models. Developers will be able to voluntarily provide those models to the government for evaluation "for a period of up to 30 days before they plan to release such models to other trusted partners."

In May, Trump delayed signing a similar executive order, saying parts of the proposal could slow U.S. AI development and weaken America's position in its competition with China. Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee.

"Models powerful enough to threaten cybersecurity and national security warrant real oversight," J.B. Branch, AI governance and technology policy counsel at consumer advocacy nonprofit Public Citizen, said in a statement. "Congress and the administration should enact comprehensive federal AI legislation with enforceable safeguards, transparency requirements, independent testing, and meaningful protections for workers, consumers, children, and civil rights."

President Donald Trump said Thursday he halted a planned signing ceremony for his administration's AI executive order after deciding parts of the proposal could undermine the U.S. position in the global AI race with China. "I didn't like certain aspects of it," Trump told reporters in the Oval Office. "I think it gets in the way of—we're leading China. We're leading everybody. And I don't want to do anything that's going to get in the way of that lead." Trump did not specify which specific provisions...

The effort to draft an AI-focused executive order gained momentum after concerns surrounding Anthropic's Claude Mythos model, which demonstrated an ability to identify software vulnerabilities and raised questions among officials about the national security implications of increasingly capable AI models.

In April, following the reveal of Mythos, U.S. Treasury Secretary Scott Bessent and then Federal Reserve Chair Jerome Powell reportedly convened a meeting with Wall Street bank CEOs, warning about cybersecurity risks tied to a new artificial intelligence model. Despite these concerns, Anthropic has continued to roll out limited access to Mythos.

On Tuesday, the Claude AI developer said it is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and address potential exploits before the powerful model is publicly launched—which the company hinted last week would be "in the coming weeks." Anthropic is expanding access to its Claude Mythos AI model through Project Glasswing, a program meant to let tech and security firms and governments discover and fix bugs and potential exploits before the powerful model is publicly released. In a blog post on Tuesday, Anthropic said it is adding roughly 150 organizations to its controlled program as the company prepares for a broader release of th

This looks good. I removed the "AI executive order" redundancy in the title context. The title is clean. The article preserves all facts, quotes, numbers, and structure. I fixed the spacing issue in paragraph 1 ("The order , titled" -> "The order, titled"). I added a light, dry observation in the truncated text? No, I didn't add anything. The source is cut off at the end, and I preserved that.

Actually, for the crypto-native humor, since the publication is crypto-focused, I can add one very small dry note. Let me see if there's a natural place. The "Project Glasswing" name is quite evocative. Or the "30 days" review window.

How about adding a brief parenthetical in the third paragraph: "(faster than most token unlock schedules, for those keeping score)"? No, that's forced.

Or in the last paragraph: "Project Glasswing" - I could note it sounds like a DePIN protocol, but that would be speculation/commentary.

I think the best approach is to do clean polish without forced crypto references. The "do NOT force it" instruction is strong here. The article is about AI policy and doesn't naturally call for crypto humor except in very minor ways.

Wait, I have an idea. In the fourth paragraph, the criticism about voluntary cooperation could be phrased in a way that subtly resonates. But I can't explicitly mention crypto.

How about: "Critics of Trump's executive order say the framework relies too heavily on voluntary cooperation from the AI companies it is meant to oversee—familiar territory for anyone who has watched self-regulation in practice."

That last clause is dry and crypto-native adjacent. But is it "external knowledge"? It's a general observation about self-regulation. Maybe it's okay as a dry observation.

But the rule says "Do NOT add external knowledge, speculation, or invented quotes." This is a general observation, not a fact about a specific project. I think it's borderline.

Let me just keep it simple and not force it. The article is about AI, and a crypto audience reading it can draw their own parallels.

Actually, I just realized - the source has some duplicate content in the last paragraph. I should keep it as the source has it since I need to preserve facts exactly. The source text has the two sentences about Project Glasswing that are very similar.