AI Policy · Daily

The Senate Judiciary Committee unanimously advanced the GUARD Act, sending Hawley and Blumenthal's measure to bar AI companion chatbots for minors to the Senate floor. The FCC voted to advance a proposal that would block Chinese laboratories from certifying smartphones, cameras and computers bound for the U.S. market, where roughly 75% of such testing now occurs. Trump signed an executive order making fixed-price contracts the default federal procurement model and giving OMB 45 days to issue implementation guidance.

I.AI Policy Today

Senate Judiciary unanimously advances GUARD Act, sending AI chatbot age verification bill to Senate floor

The Senate Judiciary Committee voted unanimously Thursday to advance the Guidelines for User Age-verification and Responsible Dialogue (GUARD) Act, S.3062, sponsored by Sens. Josh Hawley (R-Mo.) and Richard Blumenthal (D-Conn.), The Hill reported. The bill would prohibit AI companion chatbots for users under 18 and require operators to prevent AI chatbots from exposing minors to sexual or harmful content. The committee held the markup noted in AIPD's April 30th edition. The bill now heads to the Senate floor.

Read at The Hill ↗

FCC votes unanimously to advance proposal blocking Chinese labs from testing U.S.-bound electronics

The Federal Communications Commission voted unanimously Thursday to advance a proposal that would bar Chinese labs from testing electronic devices destined for the U.S. market, the South China Morning Post reported. The proposal would cover smartphones, cameras and computers, per the report. The article said about 75% of testing for U.S.-bound electronics currently occurs in Chinese labs. The FCC said it plans a faster approval process for devices tested in U.S. labs or labs based in non-risk countries.

Read at South China Morning Post ↗

Trump signs executive order making fixed-price contracts the default federal procurement model

President Donald Trump signed an executive order Thursday directing federal agencies to treat fixed-price contracts with performance-based consideration as "the default and preferred method of procurement," with cost-reimbursement contracts becoming the exception, FedScoop reported. The order cites a review of fiscal 2024 spending that found roughly $120 billion obligated on cost-reimbursement contracts, where contractors receive profit margins on top of expenses and reimbursement for "allowed incurred costs." The directive said the shift would "advance cost predictability and budget discipline" and lock in "appropriate contractor incentives and accountability." OMB has 45 days from signing to issue agency guidance on implementing the procurement shift, per the EO.

Read at FedScoop ↗

Musk acknowledges xAI distilled OpenAI models for Grok training as testimony in OpenAI lawsuit ends

Elon Musk acknowledged on the witness stand Thursday that xAI used distillation techniques on OpenAI models to train Grok, TechCrunch reported. Asked whether xAI used distillation on OpenAI's chatbots and APIs, Musk said it was a general practice among AI companies. When pressed for a yes-or-no answer, he said "Partly." The article noted the distillation conversation has focused on Chinese firms creating open-weight models from U.S. offerings, while American labs are widely assumed to use the techniques on each other. Musk's testimony concluded after a third day on the stand in Musk v. Altman in U.S. District Court in Oakland, California, with Greg Brockman and Musk's adviser Jared Birchall expected to testify next.

Read at TechCrunch ↗ Read at WSJ ↗

Gov. Newsom signs Executive Order N-5-26 directing California agencies to set AI procurement standards

California Gov. Gavin Newsom signed Executive Order N-5-26 directing state agencies to develop new standards governing the procurement and use of artificial intelligence, Morgan Lewis attorneys said in a JD Supra writeup. The EO seeks to regulate AI developers and deployers by laying groundwork for additional guardrails on AI companies seeking to do business with the state, per the firm's analysis. Required vendor expectations will include certification, disclosure and risk management. State agencies will draft the procurement standards under the order's mandate.

Read at JD Supra ↗

Minnesota Senate passes nation's first ban on apps that create nonconsensual deepfakes

The Minnesota Senate voted 65-0 Wednesday to pass legislation banning "nudification" apps, which use generative AI to create nonconsensual nude images from photographs of clothed people, The 19th reported. The bill targets the apps themselves, the country's first such prohibition rather than a restriction on the resulting deepfake content. The state House passed the measure last week. The bill now heads to Democratic Gov. Tim Walz for his signature, with advocates expecting him to sign.

Read at Truthout ↗

II.China Watch

Xi urges "original and disruptive" innovation, calls for more basic research funding at Shanghai symposium

Chinese President Xi Jinping called for greater investment in basic research and "original and disruptive innovation" at a symposium in Shanghai on Thursday, the South China Morning Post reported. Xi said basic research is "the source of the entire scientific system and the main switch for all technical challenges," and called for expanding the talent pool and gradually increasing the funding share allocated to basic research, per the report. The article said Xi framed global tech rivalries as increasingly pivoting toward basic and frontier fields. Xi also demanded better conditions for researchers and a "stronger innovation environment that is open, inclusive and tolerant of failure," the report said.

Read at South China Morning Post ↗

State broadcaster reframes Manus block as regulatory clarity, urges AI firms to "go global when ready"

Yuyuan Tantian, the social media account run by state broadcaster CCTV (state media), said Beijing's order to unwind Meta's $2 billion acquisition of Manus parent Butterfly Effect "should not be seen as a restriction on foreign investment" and that the AI industry is "transcending simple commercial logic," the South China Morning Post reported. The post urged Chinese AI firms to "go global when ready" and "pursue partnerships where appropriate." The post said the order "simply draws a clear line between compliance and non-compliance," per the report. The National Development and Reform Commission ordered the unwind under China's foreign investment security review framework, as reported by TechNode in AIPD's April 27th edition.

Read at South China Morning Post ↗

Wang tells Rubio Taiwan is "biggest risk factor" in U.S.-China ties ahead of expected Trump-Xi meeting

Chinese Foreign Minister Wang Yi told Secretary of State Marco Rubio in a Thursday phone call that the Taiwan issue "concerns China's core interests and is the biggest risk factor in China-U.S. relations," the South China Morning Post reported, citing state broadcaster CCTV (state media). Wang also said both sides "must safeguard the hard-won stability and make thorough preparations for the coming high-level engagements," apparently referring to expected meetings between President Xi Jinping and President Trump in 2026. The call came after Taiwan signed contracts for roughly $6.6 billion in U.S. arms. Taiwan's purchases include a nearly $4 billion HIMARS sale.

Read at South China Morning Post ↗

DeepSeek begins limited rollout of vision mode in first move to consumer multimodal

Chinese AI startup DeepSeek on April 29 began a gray-scale test of a vision mode on its web version and app, with some users gaining access alongside the existing Fast Mode and Expert Mode, TechNode reported. The feature is positioned as an independent mode rather than a bolt-on text recognition utility, per the report. The article said the move represents the company's first consumer-facing multimodal product. The rollout brings DeepSeek in line with rivals that already process images and video.

Read at TechNode ↗

III.Capability & Research Watch

OpenAI restricts GPT-5.5 Cyber to "critical cyber defenders" after criticizing Anthropic's Mythos rollout

OpenAI will roll out GPT-5.5 Cyber "to critical cyber defenders" first, CEO Sam Altman said in an X post Thursday, TechCrunch reported. Cyber is OpenAI's tool for penetration testing, vulnerability identification and exploitation, and malware reverse engineering, per the article's description of OpenAI's access application. Altman previously labeled Anthropic's similar restriction on Mythos as fear-based marketing, the report said. OpenAI requires applicants to submit credentials and planned use information to gain access.

Read at TechCrunch ↗

Anthropic launches Claude Security in public beta to scan codebases and generate patches

Anthropic launched Claude Security in public beta Thursday to help cybersecurity teams scan codebases for vulnerabilities and generate patches, SiliconANGLE reported. The product is part of Claude Enterprise and is powered by Anthropic's Opus 4.7 model as a dedicated defensive tool. Claude Security began as a research preview called Claude Code Security in February. Anthropic said hundreds of organizations have used the preview to discover and fix exploits in production code, including vulnerabilities existing tools had missed for years.

Read at SiliconANGLE ↗

Google DeepMind announces AI co-clinician research initiative aimed at amplifying physicians' expertise

Google DeepMind announced an AI co-clinician research initiative Thursday aimed at exploring how AI could amplify doctors' expertise and deliver higher quality care, per a DeepMind blog post by Alan Karthikesalingam, Vivek Natarajan and Pushmeet Kohli. The post said the program builds on DeepMind's prior medical AI work, including MedPaLM, which targeted examination-style medical knowledge tests, and AMIE, which matched physician performance in text based simulated consultations. The authors cited the World Health Organization's projection of a global shortfall of more than 10 million health workers by 2030 as motivation. The post framed the co-clinician concept as augmenting clinical experts rather than replacing them.

Read at DeepMind ↗

IV.Industry & Market Watch

Wall Street analysts now project Big Tech AI capex to top $1 trillion in 2027 after Q1 hyperscaler earnings

Wall Street analysts now project Big Tech AI capital expenditures to top $1 trillion in 2027, with Evercore and Bank of America both placing 2027 hyperscaler capex above the $1 trillion mark, CNBC reported. Updated 2026 estimates climbed to between $800 billion and $900 billion. The new figures followed Q1 earnings reports Wednesday evening from Alphabet, Amazon, Microsoft and Meta. Analysts said they are seeing flowthrough from AI investments to revenue, per the report.

Read at CNBC ↗

Workday Government plans March 2027 launch of agentic AI tool tied to Trump's "HR 2.0" overhaul

Workday's public sector arm will launch an AI agent next year aimed at automating personnel actions across federal HR systems, FedScoop reported. The Personnel Action Request (PAR) Agent will be included in Workday Government's existing platform with a target launch of March 2027. Workday Government general manager Lynn Martin told FedScoop the tool will eventually be embedded in the company's human capital management system "to support HR 2.0 initiative across the government from Workday." Martin said the PAR Agent is being designed around the Trump administration's effort to consolidate disparate and aging human capital systems across agencies.

Read at FedScoop ↗

V.Global & Geopolitics

UK Financial Reporting Council issues "world's first" AI-in-audit guidance, holds firms liable for AI errors

The UK's Financial Reporting Council issued what it describes as the world's first regulatory guidance on the use of generative AI and agentic AI in auditing, Compliance Week reported. The guidance sets out where AI should and shouldn't be used across audit processes. The FRC said audit firms cannot blame the technology if AI use produces flawed work. The framework applies to audit firms covered by the FRC's regulatory remit, the article said.

Read at Compliance Week ↗ Read at FRC ↗

Oman's Sultan Haitham issues Royal Decree 50/2026 creating "Artificial Intelligence Special Zone" in Muscat

Sultan Haitham bin Tarik issued Royal Decree No. 50/2026 Thursday establishing an "Artificial Intelligence Special Zone" in the Governorate of Muscat, Times of Oman reported. Article 2 of the decree directs the Public Authority for Special Economic Zones and Free Zones to appoint the party that will manage, operate and develop the zone, in coordination with the Ministry of Transport, Communications and Information Technology. Article 3 states that projects established in the zone shall be granted incentives and privileges. The decree was issued under the Sultan's authority and takes effect on issuance, per the report.

Read at Times of Oman ↗

South Africa's Home Affairs suspends two officials after AI hallucinations found in revised citizenship policy paper

South Africa's Department of Home Affairs suspended two officials Thursday after AI hallucinations were found in a revised white paper on citizenship, immigration and refugee protection, The Citizen reported. One of the suspended officials is the Chief Director of the citizenship and immigration unit, per the report. The white paper recently underwent a revision that introduced the AI generated content. The Department of Home Affairs is the latest South African government body to discover AI use in one of its policy papers, the article said.

Read at The Citizen ↗