Trump Postpones Signing of AI Executive Order Citing Concerns Over Competitiveness
On May 21, 2026, President Donald Trump postponed the signing of an executive order on artificial intelligence, stating he was concerned it could hinder U.S. leadership in the technology sector. Trump said he decided to delay the signing hours before it was scheduled, citing dissatisfaction with aspects of the document. The order, which had been in development for weeks, aimed to strengthen federal oversight of advanced AI models, particularly in cybersecurity and pre-deployment testing. It was expected to involve collaboration with major AI companies and federal agencies including CISA, ONCD, and OSTP. The proposed action followed growing concerns after Anthropic's Mythos model demonstrated capabilities to identify critical system vulnerabilities. While Trump emphasized protecting American competitiveness, particularly against China, administration officials had previously stressed the importance of data privacy and public safety in AI development.
NBC News provides the most complete and contextually rich account, while ABC News and New York Post offer increasingly narrow and decontextualized versions of the same event.
- ✓ President Trump postponed the signing of an executive order on artificial intelligence on May 21, 2026.
- ✓ The announcement was made by Trump during an unrelated event in the Oval Office.
- ✓ Trump cited concerns that the order could harm U.S. competitiveness in AI, particularly relative to China.
- ✓ Trump stated he 'didn’t like' aspects of the order’s content.
Context and significance of the order
Portrays the order as potentially harmful to industry competitiveness but provides no detail on its content.
Describes the order as 'landmark' with specific goals: cybersecurity defenses and voluntary testing of frontier AI models.
Offers no information about the order’s substance or implications.
Reason for postponement
Focuses on Trump not liking the text and wanting to preserve U.S. lead.
Adds that Trump believed the order 'could have been a blocker' and includes context about cybersecurity risks motivating the order.
Reduces rationale to Trump’s personal dislike of 'certain aspects'.
Background and urgency
No mention of external events or technological developments.
Links the order to Anthropic’s Mythos model and its demonstrated ability to find critical vulnerabilities.
No contextual information provided.
Policy scope and implementation
No details on agencies involved or order structure.
Specifies agencies (CISA, ONCD, OSTP) and outlines two-part structure: cybersecurity and AI model testing.
No policy details included.
ABC News
Framing: The event is framed as a presidential decision driven by economic and technological competitiveness concerns. The focus is on Trump’s personal reaction to the content of the order and his desire to maintain U.S. leadership in AI.
Tone: Neutral to slightly informal, with an emphasis on immediacy and presidential agency.
Framing by Emphasis: Emphasizes Trump’s concern about 'dull[ing] America’s edge' rather than regulatory intent or public safety.
"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"
Vague Attribution: No sourcing for claims about the order’s development or content; relies solely on Trump’s statements.
"He didn’t like what he saw in the order’s text"
Omission: Does not mention any context for why the order was created, such as cybersecurity risks or recent AI developments like Mythos.
NBC News
Framing: The event is framed as a significant policy reversal involving a 'landmark' executive order, with attention to its intended regulatory scope, development context, and national security implications.
Tone: Analytical and detailed, with a focus on policy substance and background context.
Comprehensive Sourcing: Cites 'two people familiar with the draft' and includes statements from Vice President JD Vance.
"according to two people familiar with the draft"
Narrative Framing: Places the delay in the context of a broader timeline involving Anthropic’s Mythos model and rising cybersecurity concerns.
"The order’s planned debut came less than two months after AI company Anthropic’s Mythos Preview model demonstrated the novel ability..."
Balanced Reporting: Presents both the competitiveness argument (via Trump) and the rationale for regulation (via Vance and context on vulnerabilities).
"Right now, we’re working in a collaborative way with the technology companies..."
Proper Attribution: Clearly attributes information to specific sources, including officials and unnamed insiders.
"according to the sources"
New York Post
Framing: The event is framed narrowly as a personal decision by Trump based on subjective dissatisfaction with the document’s details.
Tone: Concise and minimal, bordering on dismissive in its brevity.
Cherry-Picking: Focuses exclusively on Trump’s statement that he 'didn’t like' the order, without exploring content, intent, or consequences.
"I didn’t like certain aspects of it, I postponed it"
Omission: Provides no background on the order’s purpose, development, or potential impact; omits mention of cybersecurity threats or interagency roles.
Editorializing: Includes promotional content ('Follow The Post’s live coverage...') that suggests a media brand presence, potentially influencing framing.
"Follow The Post’s live coverage of President Trump and national politics for the latest news and analysis"
NBC News
Provides the most comprehensive coverage: includes background context, sourcing, policy details, interagency roles, and political framing from multiple actors.
ABC News
Offers basic facts and direct quotes but lacks depth on the order’s content or rationale beyond Trump’s competitiveness argument.
New York Post
Most minimal in content; presents only a fragment of Trump’s statement with no expansion or context.
Trump scraps signing of landmark executive order regulating AI
Trump says he postponed signing AI order because he didn’t ‘like’ it
Trump says he's postponing signing an executive order on AI out of concern it would hurt AI industry