The Cosmos Ecosystem UX Chain will be gradually paused and is expected to fully shut down by May 15.

Beating AI News Flash: OpenAI is pushing for the establishment of unified mandatory national AI safety rules in the U.S., calling on Congress to directly regulate the most cutting-edge AI firms. The proposed rules cover unified testing, independent security assessments, cybersecurity requirements, and reporting of major incidents. OpenAI hopes the government will further set unified safety thresholds, clarify when AI development must slow down or even halt, and advocate for similar standards to be adopted by other countries. This regulatory framework targets only a small number of major companies developing the most powerful models, excluding regular startups and small-scale developers. Additionally, OpenAI announced its support for four AI safety bills in California, some of which it had not backed previously. The company noted that its stance has shifted amid the rapid advancement of model capabilities.

Beating AI Express News reports that tech industry user Leo posted on X, stating that per his knowledge, OpenAI is very close to verifying the Hodge Conjecture, while either OpenAI or Anthropic may also be near solving the Birch-Swinnerton-Dyer Conjecture. The user stressed that the relevant information is currently just a rumor with unconfirmed authenticity, but if proven true, it would mean three Millennium Prize Problems could be solved in just one month. Both the Hodge Conjecture and the Birch-Swinnerton-Dyer Conjecture remain unsolved Millennium Prize Problems. Prior AI breakthroughs in mathematics have drawn market attention, and prediction markets now rank the two conjectures as the next Millennium Prize Problems that AI may solve.

Beating AI News Flash: OpenAI has brought former alignment research lead Paul Christiano back to its board. Christiano led alignment research at OpenAI from 2017 to 2021 and contributed to the foundational work of RLHF. After leaving, he founded the Alignment Research Center. This time, he is joining the board of the OpenAI Foundation and its Safety & Security Committee, while also joining the board of the for-profit entity OpenAI Group as an observer. In his first public statement after taking office, he directly named OpenAI. Christiano warned that automated AI development could soon lead to rapid acceleration in model capabilities. If superintelligence is developed before the alignment problem is solved, humans could lose control permanently, resulting in "most people likely dying". He also stated that the entire AI industry, including OpenAI, has not yet reduced this risk to an acceptable level.

According to Lookonchain monitoring, the mysterious whale address 0x6436 purchased another 308,569 HYPE tokens today, valued at approximately $26 million. As of now, this address has accumulated around 4.05 million HYPE tokens for a total value of about $322 million, and has staked all of its HYPE holdings.

Beating AI News Flash, according to a report from Axios, a Republican-led subcommittee on disaster management of the U.S. Senate is investigating OpenAI’s handling of the July Hugging Face security breach, a move coming as Capitol Hill’s concerns over AI existential risk are rapidly escalating. Earlier, multiple researchers from Anthropic and OpenAI had publicly issued warnings. Missouri Republican Senator Josh Hawley, chair of the Senate Homeland Security and Governmental Affairs Committee’s Subcommittee on Disaster Management, wrote in a letter to OpenAI CEO Sam Altman: "As you know, an increasing number of AI experts in the public domain are warning about AI existential risks." He added: "This week alone, three Anthropic researchers publicly stated that the probability of AI killing all humanity within the next decade exceeds 10%." Hawley noted that the investigation is in response to OpenAI’s recently released internal investigation findings. The Hugging Face breach marks a turning point in AI history, prompting OpenAI to slow the release of its own models and alert the industry to AI-driven cyberattacks.

At the Republican midterm election conference, Trump made a bold promise: if the GOP retains control of the House of Representatives and Senate in the November midterms, he will distribute a $5,000 "Trump Dividend" to every adult U.S. citizen. The plan is estimated to cost over $1 trillion, though Trump did not specify its funding source, and its implementation would require congressional approval. This rare promise underscores the intense midterm election pressure facing Trump and the Republican Party. Trump’s approval rating remains under pressure, voter discontent over living costs and the Iran conflict is rising, and multiple Republican candidates in tight races even skipped the conference; the GOP currently needs to hold onto its narrow majority in the House, while Senate seats are also competitive. In his speech, Trump unusually asked voters to "pretend I’m on the ballot" and acknowledged that the president’s party typically loses congressional seats in midterm elections. He also pledged to devote more time to helping Republican candidates campaign, seeking to reposition himself as the core mobilizing force for the November elections. Meanwhile, the U.S. national debt has surpassed $40 trillion, raising questions about the fiscal viability of the $5,000 dividend plan. The Republican Party has also begun seeking adjustments to the plan, with Senator Ted Cruz suggesting it could be structured as a tax refund with work requirements.