Trump's Super Intelligence Summit, AI Safety Accord, GDP Beats, Midterm Predictions

David Sacks described President Trump’s White House summit as the first gathering to span the AI industry’s data-center, chip, hyperscaler, and frontier-model layers. Trump opened by saying, “Whoever wins superintelligence wins,” and urged companies to show AI’s benefits, support local communities, and maintain checks and balances. Six major frontier-model companies signed the White House Accord on Super Intelligence, which Sacks said commits them to take responsibility for development and establish internal controls subject to external audits, with reports going to an independent board committee.

The accord’s safety language was expanded to include internal controls, outside audits, and board oversight alongside a continuing safety working group. Sacks said the commitments could create enforceable obligations through board fiduciary duties and companies’ public statements, even though signing was voluntary. He presented the agreement as a workable alternative to pausing frontier development, arguing that a pause could allow China to pull ahead. EY and 8090 were building audit infrastructure to trace policies to risks and produce evidence for lawyers, regulators, customers, or courts; Sacks said the controls could be implemented quickly without new legislation or an international agreement.

Friedberg argued that open-source and open-weight models, including models that run locally on small machines, make centralized control of AI development impractical. He expected rising risks to expand cyber-defense systems, investment, and board-level budgets, and predicted that within 12–18 months the debate would shift toward data-center access and allocation of GPUs, chips, and servers. Regulators would likely have a role, he said, and the compute used for defense relative to attackers’ compute could become a key measure of cyber-defense capacity. Speakers also cast data centers and domestic electricity production as national-security priorities. One cited China’s grid doubling roughly every decade while the U.S. grid had been roughly flat for 25 years, and favored stronger privacy laws to address surveillance concerns rather than blaming data centers.

Speakers said the summit was a first step but did not sufficiently address potential job losses or how AI could benefit education, housing, and health care; one recommended inviting teachers and health-care representatives to future events. They disagreed over public anxiety: one called doomerism unsupported by data and politically useful, while another said concerns such as flight attendants’ fears of job loss still need answers. Supporters argued that open models could spread benefits without companies charging a toll, but that warnings about AI were drowning out the case for openness and economic upside. One speaker said available data suggested more work and higher earnings over the next four to five years.

Economic figures cited were stronger than expected: Q2 GDP was revised to 2.2% from 1.5%, and Q1 to 2.5% from 2.1%; Atlanta Fed GDPNow estimated Q3 growth at about 3.7%. August payrolls rose by 162,000 against 55,000 expected, while June and July were revised up by 55,000 combined. Unemployment was 4.1%, labor-force participation 61.6%, and core PCE inflation 3.0% against 3.3% expected. One speaker said inflation remained too high but was cooling and pointed to expanding manufacturing surveys as evidence of reindustrialization. Another disputed the “K-shaped recovery” narrative, citing median household income near $90,000 in 2025 dollars, a 10.2% poverty rate described as a historic low, and 3.1% growth in after-tax household income in 2025; a co-panelist countered that people did not feel the gains.

The economic outlook remained contested. Speakers flagged diesel prices and rising short-term Treasury rates as risks to consumer costs and borrowing, while disagreeing on whether easing fuel costs could bring inflation down and enable rate cuts. One said higher diesel costs could feed into goods prices and that GDP could reach 5% if diesel fell; another said gas prices were up 50% since the Iran War began and were already hurting consumers. Midterm predictions also diverged: Jason forecast a Democratic sweep, citing Polymarket odds of 64% for Democrats to win both chambers, 63% for the Senate, and 93% for the House. Chamath saw the Senate as a toss-up Republicans could retain and expected Democrats might narrowly win the House; another speaker expected Republicans to keep the Senate. Speakers said the result could turn on whether Republicans emphasized economic gains or Democrats kept attention on gas prices and AI fears, and warned that polls can miss. Friedberg said the 2026–2028 period would also be shaped by Democratic movement left and Republican fractures around populism, while looming Social Security, interest-rate, and affordability pressures could force unpopular choices on taxes, spending cuts, or entitlements.