Monday Morning News Roundup 09/28/2026

• A summit is a start, not a result
• Iran: What would end the war?
• Voting rights cannot depend on a database guess
• Who built the government’s software?
• Trust has to be earned

There was a lot of ceremony in Washington this week. President Donald Trump welcomed Chinese President Xi Jinping, and the two governments announced plans to talk about artificial intelligence, keep military communication channels open, and ease tariffs on a limited set of goods. In a world where a mistake between the United States and China could become a catastrophe, I’m glad they are talking.

But I kept thinking about a less ceremonial question: What happens between the announcement and the moment one of these systems fails?

That question applies to more than diplomacy. It applies to the federal database states are using to check voter citizenship. It applies to software sold to agencies entrusted with sensitive information. And it applies to a war in which an offer to reopen a major shipping lane was rejected while threats to bases and neighboring countries remain live concerns.

I don’t think the answer is to reject every tool, every security measure, or every negotiation. It is to stop mistaking a reassuring label—safety, verification, national security—for proof that the people in charge have earned our trust.

§A summit is a start, not a result

After Xi’s visit to Washington, the United States and China announced tariff cuts on roughly $30 billion in non-sensitive goods in each direction. Examples cited in the reporting included American agricultural goods, wood, and cosmetics, and U.S. imports such as small appliances, toys, and decorations. They also announced an AI dialogue and a channel for communicating about AI-related incidents, alongside plans to strengthen military crisis communications. The next AI discussions are scheduled for November.

Those are worthwhile steps. A channel that lets officials speak when an AI-related incident or military encounter threatens to spiral is better than forcing them to improvise in a crisis. A limited tariff reduction is better than insisting every ordinary product must become a weapon in a trade war.

But let’s keep the scale straight. This is not a comprehensive trade deal, an AI safety treaty, or a resolution of the countries’ disputes over Taiwan, technology, or Iran. Reuters described the summit as heavy on symbolism and light on major breakthroughs. The subsequent announcements offered working-level mechanisms and narrower economic steps, not a settlement of the relationship.

There is a telling detail in the AI announcement: the White House said the leaders agreed to use the term “super intelligence” instead of “artificial intelligence”. Call it whatever you want. The name is not a safety standard. The useful questions are whether the two sides will share timely incident reports, establish reliable emergency contacts, test those channels, and actually use them when something goes wrong.

I also want to know what Americans get from the tariff changes beyond a headline number. Which rates will change? When? For how long? Will businesses and consumers see the difference? If this is a genuine economic gain, we should be able to measure it—not merely admire the state dinner.

And Taiwan should not disappear behind the pageantry. Xi pressed his position on Taiwan during the visit, according to Reuters. A calmer relationship between Washington and Beijing is welcome, but stability cannot mean making consequential decisions about Taiwan’s future through vague private understandings that the people affected cannot evaluate.

The best outcome here would be unglamorous: durable channels, clear rules, fewer opportunities for an accident to become a confrontation, and public accounting of what was actually agreed. The danger is confusing a photograph of two leaders getting along with evidence that the hard problems have been solved.

§Iran: What would end the war?

The contrast with Iran is hard to miss. The United States and China say they agree that Iran should not develop nuclear weapons and that international waterways should not be subject to transit tolls. Meanwhile, the Strait of Hormuz remains a point of confrontation.

According to the Associated Press, Iran proposed reopening the strait and resuming nuclear talks within seven days if the United States lifted its blockade of Iranian ports, waived sanctions on Iranian oil sales, and observed a ceasefire that included Lebanon. Trump rejected the proposal, saying the deal was not acceptable.

That does not mean Iran offered terms Washington was obliged to accept. Each condition has consequences, and an announcement is not the same as a trustworthy agreement. Iran’s threats to vessels and its effective closure of the strait are serious acts with costs far beyond the region.

But rejecting a proposal should be followed by more than a declaration that the other side is losing. Which terms were unacceptable? What counteroffer would the United States consider? Is the goal a verifiable nuclear agreement, a reopened waterway, a broader ceasefire—or something else? If the administration wants the public to support continued military pressure, it owes the public a clear description of what that pressure is meant to achieve.

The risk of the conflict widening is not hypothetical. Saudi Arabia said it intercepted a missile it attributed to Yemen’s Iran-backed Houthis near Khamis Mushait. The AP could not independently verify the claim, and there was no immediate claim of responsibility from the Houthis. Those qualifications matter; so does the fact that regional governments are responding to reported threats while the war continues.

The war’s reach was also part of the context for Sunday’s arrests near RAF Fairford, a British air base used by U.S. forces in operations against Iran. British police arrested five men on suspicion of explosives offenses and preparation of a terrorist act, evacuated about 85 nearby households as a precaution, and said the incident was contained. Authorities had not established a connection to Iran or publicly confirmed a motive. A base’s role in the war makes the story especially concerning; it does not prove who the suspects are or why they allegedly acted.

That distinction is essential. Take a potential threat to service members and neighbors seriously. Support a thorough investigation. Do not turn an arrest into a completed account of a plot—or a claim about Iranian involvement—before investigators establish the facts.

For me, the central question remains: what would end this war? If diplomacy is possible, spell out the terms. If military action is necessary, state the objective, the limits, and the authority for it. “We’re winning” is not an exit strategy.

§Voting rights cannot depend on a database guess

Back home, the Supreme Court has temporarily cleared the way for states to use the Trump administration’s expanded Systematic Alien Verification for Entitlements database, known as SAVE, to help check the citizenship status of people on voter rolls. The emergency order paused a lower court’s restriction; it did not settle every legal question about the program or require every state to use it. Three justices dissented.

The administration argues that states need access to federal information to prevent noncitizens from voting. I think eligible voters and election officials alike deserve accurate rolls. But a database match is not a verdict, and this system’s error rate has not been publicly established. Reuters reported that some people flagged as potential noncitizens have already had their registrations canceled. The AP documented an eligible naturalized citizen in Texas whose registration was temporarily canceled while he obtained a new passport.

Imagine learning you had been flagged just before an election—not because you committed fraud, but because a government record failed to reflect that you became a citizen years ago. Now imagine being told the burden is yours to fix it in time.

That is the flaw in treating mass verification as if it were automatically precise. The Supreme Court majority said federal law authorizes information-sharing about citizenship and immigration status. Justice Ketanji Brown Jackson, joined by the other dissenters, argued that the risk of burdening or disenfranchising even a small number of lawful voters was too great.

There are limits on what Friday’s order can do immediately: participation by states is voluntary, and existing law restricts most systematic voter-roll removals close to an election. But those limits do not make erroneous flags harmless. An individual voter can still be forced to spend time, money, and energy proving a status the government should have checked carefully before taking action.

My standard is simple. Check eligibility lawfully. Treat a database result as a lead, not a finding. Give people clear notice, access to the evidence, a realistic opportunity to correct errors, and a reliable way to keep voting while a mistake is resolved. Publish the error rate. Audit the system independently. A government that cannot explain how often its tool is wrong has no business asking citizens to trust it with their right to vote.

§Who built the government’s software?

The week’s most revealing technology story may be the least flashy. Federal prosecutors have charged Lee Reiber and Oleg Davydov with conspiracy to commit wire fraud. Reiber is the chief executive of Virginia-based digital-forensics firm Oxygen Forensics; Davydov is a Russian national. The Justice Department alleges they concealed Russian ownership and control of the company, as well as the fact that its software was developed in Russia, while pursuing U.S. government business. Agencies identified in the complaint include the Secret Service, Homeland Security Investigations, the Department of Homeland Security’s inspector general, and the Department of War.

This is a criminal allegation, not a conviction. The defendants are presumed innocent. And there is a particularly important limit to the story: the complaint does not allege that the software contained malicious code or that anyone used it to gain unauthorized access to a customer’s systems or data.

That is precisely why the story deserves careful attention rather than a sensational headline about proven espionage. If the government’s allegations hold up, the failure was at the point where officials decide whom to trust. Who owns a contractor? Who controls the company? Where is the software built? Who can influence updates and development? Are certifications checked, or merely filed?

The Justice Department says Oxygen Forensics won a five-year contract with the Secret Service’s National Computer Forensics Institute in 2024 after certifications about ownership that prosecutors say were false. It also alleges that the Russian owners continued to make significant corporate decisions. Those claims need to be tested in court. In the meantime, federal agencies should be able to explain how they assessed the vendor and what steps they took once concerns emerged.

Digital-forensics tools can handle sensitive material. The public has a reasonable interest in whether agencies independently verify the provenance and control of such tools. But there is no need to invent a breach to make that case. The alleged deception, if proven, is serious enough on its own.

There is a lesson here for the AI and surveillance debates, too. A system can have a domestic office, a polished sales pitch, and government customers—and still leave unanswered questions about who built it and who controls it. Procurement is a security decision. Verification has to happen before a vendor is trusted, not only after prosecutors file charges.

§Trust has to be earned

These stories are not identical. A summit with China is not a voter database. A criminal complaint is not proof of a cyber intrusion. Arrests near an air base do not establish an Iranian plot. A rejected diplomatic proposal is not, on its own, proof that diplomacy is impossible. The distinctions are the point.

We can welcome U.S.–China crisis channels without pretending the summit settled the rivalry. We can demand accurate voter rolls without allowing faulty records to cancel citizens’ registrations. We can investigate alleged deception by a software vendor without claiming a hack no one has alleged. We can protect an air base without guessing at a motive. And we can oppose attacks on shipping while insisting our government explain how it intends to end a war.

The standard I want from people in power is not complicated: tell us what you know, tell us what you do not know, explain who can challenge a mistake, and show us where the limits are. That is how institutions earn trust. Asking for trust while withholding the means to verify it is something else entirely.