I ran across an interesting post by Kevin Bass about Anthropic, AI safety, and the network of organizations involved in evaluating the risks posed by AI.

I don’t endorse everything in the post. In particular I don’t think it is necessary to assume corruption, conspiracy, or even bad faith to recognize the problem he identifies.

The problem is incentives.

Anthropic has been among the loudest voices warning that advanced AI may pose extraordinary risks. It has also advocated government regulation of frontier models, including independent evaluations intended to determine whether those models pose unacceptable risks.

Let’s assume they’re right.

Indeed, let’s go considerably farther than that. Let’s give Anthropic the keys.

If Anthropic really has the expertise to determine what constitutes a dangerous AI model, put it in charge of developing the standards. Give it an important role in evaluating models. Give its recommendations substantial regulatory force.

There should be one condition.

Anthropic can’t profit from it.

I don’t mean that figuratively. Congress has considerable power to establish the conditions under which governmental authority may be exercised. We already have conflict-of-interest laws, disclosure requirements, divestiture requirements, procurement rules, and restrictions intended to prevent people from using governmental authority for their own financial benefit.

Use them.

Construct whatever legal mechanism is necessary so that Anthropic, its principals, and those exercising the delegated authority cannot become richer because of the regulatory regime they devise. That might require divestiture. It might require some sort of regulated return. It might require a special corporate structure. Those are details for lawyers to work out.

The principle is simple enough: you can have the keys or you can have the money. You can’t have both.

There is a reason for doing that which has nothing to do with whether anyone at Anthropic is honest.

Regulation creates barriers to entry. The more elaborate the testing requirements, the more expensive compliance becomes. A company already spending billions of dollars developing frontier models is in a much better position to comply with a regulatory regime requiring billions of dollars than a prospective competitor is.

Consequently a company can advocate regulations for perfectly sincere reasons and still benefit enormously from them. The regulations may protect the public. They may also protect the incumbent.

Both things can be true at the same time.

That is why I don’t think accusations of corruption are particularly useful here. We don’t need to know what is in anybody’s heart. We can remove the conflict instead.

There is another advantage to this approach. It would provide a pretty good test of revealed preference.

If the people at Anthropic actually believe that frontier AI presents risks to human civilization sufficiently grave to justify extraordinary governmental intervention, asking them to surrender the opportunity to make extraordinary private profits from that intervention doesn’t seem unreasonable.

The greater the danger, the stronger the argument becomes.

Conversely, if the response is that giving up those profits would be intolerable, perhaps the situation is not quite as extraordinary as we have been told.

I don’t expect anything like this to happen. The major AI companies would oppose it. The investors would oppose it. Quite possibly the AI safety organizations would oppose it. There would be endless arguments that government could not afford to lose access to the expertise concentrated in the companies actually developing frontier AI.

My proposal doesn’t lose that expertise.

Use it.

Give Anthropic the keys.

Just don’t let them own the tollbooth.

A high-ranking Chinese official explains why China may cooperate in trying to control the pace of development of LLM AI and why such plans won’t succeed in one neat package. Carol Yang reports at the South China Morning Post:

China’s top intelligence official has warned of rising national security risks posed by artificial intelligence (AI), calling for robust risk prevention frameworks and stronger global governance of the emerging technology.

The call from Chen Yixin, China’s state security minister, comes as Beijing seeks to play a leading role in global AI governance while criticising US restrictions. At the Brics summit in India over the weekend, Chinese President Xi Jinping proposed an AI-powered initiative to drive new industrialisation and deepen supply chain cooperation in the emerging economies bloc.

Chen underscored Beijing’s view of the technology as a key geopolitical battleground. AI has become “a new arena for strategic rivalry among major powers”, he wrote for China Cyberspace, a journal run by internet watchdog body the Cyberspace Administration of China.

The aspects of the matter are:

1. China has reasons to fear uncontrolled AI because of regime security.
2. China also has enormous incentives to exploit AI for strategic advantage.
3. The United States faces analogous strategic incentives.
4. The technology and expertise are already too widely distributed for an agreement between two governments to control development reliably.

Consequently, even if the U. S. and China reach an agreement to slow the pace of development there are quite literally millions of people in the U. S. and China with the knowledge, training, abilities, and resources to keep right on developing it and reasons to do so.

Feel lucky, punk?

Inspired by last week’s remembrances of the attacks on 9/11, I set out to research the prevalence of terrorist attacks in the United States since then. What I learned astonished me. What surprised me was not how common or rare such attacks have been but how controversial the definition of terrorism itself remains.

There is no internationally accepted definition of terrorism. Or, rather, there are lots of definitions—there is no single accepted one.

I would define terrorism as violence or the threat of violence intended to coerce public action by frightening or intimidating a population for a political or ideological objective. That definition has several components: violence or its threat; intimidation extending beyond the immediate victims; an attempt to coerce action; a political or ideological objective; and some question of who or what may legitimately be targeted. Somehow one or more of those components has eluded agreement whenever the international community has tried to define terrorism.

One distinction seems particularly important. Political violence is not necessarily terrorism. Political assassination is not necessarily terrorism. Anti-government violence is not necessarily terrorism. Terrorism is distinguished not merely by the motive of the perpetrator but by the use of violence to intimidate a broader population in order to coerce political action.

I can’t help but wonder whether the reason a definition has eluded us is that a rigorous definition would constrain the parties doing the defining. A definition broad enough to encompass everything one side wants to condemn may also encompass actions by itself or its allies that it considers legitimate.

The moment that many of us have been dreading has arrived. Chicago Mayor Brandon Johnson has announced that he will seek re-election. ABC 7 Chicago reports:

CHICAGO (WLS) — Chicago Mayor Brandon Johnson said he will run for reelection.

The mayor has been hinting at is announcement with plans for a “special announcement” on Sunday.

Johnson posted a two-minute video ending with people chanting “four more years” on X.

He later posted: “I’m here because I’m the only candidate who’s fighting for working families. If you’re ready for a safer and more affordable Chicago, then now is our time!”

Johnson is entering an already crowded race. At least 11 candidates have already announced their intentions to run.

To place that in perspective consider the approval ratings of previous incumbent Chicago mayors when they decided to run for re-election:

Mayor	Approval rating
Richard M. Daley	41%
Rahm Emanuel	42%
Lori Lightfoot	22%

Some recent polls have shown Johnson’s approval rating in single digits. If the CTU decides to support his candidacy again of which I have little doubt, at least he’ll have that going for him.

I just finished hearing the CEO of Anthropic say the following on one of today’s “talking heads” programs:

The biggest dilemma is how we get China to agree to join the agreement on regulating the pace of AI development.

but I think he’s wrong. The dilemma that’s bigger than that is how we prevent a country we support that’s engaged in an existential conflict against a larger foe from developing their own LLM AI that breaks all the rules when it means their survival as a country.

That’s also the problem with Megan McArdle’s more modest and seemingly more doable agreement which she describes in her most recent Washington Post column:

Sure, I can imagine China and the United States coming to a rapid-fire agreement for an AI pause. But then, I can also imagine getting a MacArthur “genius grant,” which seems about as likely.

So back to my question about the murderbots. Could we at least not build the literal, physical murderbots, such as fully autonomous drones that can operate without humans in the kill chain? Controlling their development would be a very hard problem, as all arms control is. But it seems more manageable than completely stopping AI advancement.

Consider her proposal in the context of Ukraine. Should we actively discourage even prevent Ukraine from developing their own “murderbots” if they make the difference between their victory or defeat against the Russians?

The bigger dilemma may be how you construct an AI arms-control regime that remains rational to obey when one of its members believes violating the rules is necessary for national survival.

There’s an even more elementary problem. Suppose Anthropic, OpenAI, xAI, and Google agree on rules limiting what their LLM AIs may do. What prevents a well-financed startup from developing an LLM AI that ignores those rules?

If the prohibited capabilities are useful, that company’s product may be more competitive precisely because it breaks the rules. The agreement would then have accomplished something rather peculiar: it would have imposed competitive disadvantages on the companies that agreed to it while creating an opportunity for companies that did not.

Preventing that requires something considerably more ambitious than an agreement among AI companies—or even an agreement between the United States and China. It requires preventing *anyone capable of developing the technology* from defecting when there is an economic or strategic advantage in doing so.

Now consider Ukraine. If we cannot be confident that a startup will refrain from breaking the rules for money, why should we expect a country to refrain from breaking them when its leaders believe national survival is at stake?

When we woke the morning after leaving Ketchikan, the ship was already docked in Juneau. After showering and breakfasting we left the ship and spent the rest of the morning sightseeing.

Neither my wife nor I were particularly interested in the tourist-y downtown area so we immediately headed off to the Tongass National Forest. We caught our first glimpse of a glacier while walking in the Tongass National Forest near Juneau. You can see it in the photo at the top of the post.

We continued walking until we reached a very scenic waterfall.

Following that we returned to the ship, got lunch, and rested for the remainder of the day.

James Joyner has a fine round-up of posts and articles relating to the commemoration of the anniversary of the terrorist attacks on September 11, 2001. I’m going to devote this post to a few ramblings of my own on that subject.

There have been many changes as a consequence of 9/11. Many of them are institutional; fewer have improved our actual security. Key issues include

• American military presence in the Middle East
• Islamist terrorism
• Intelligence
• Airport security
• Surveillance
• Federal bureaucracy
• War powers
• Fiscal consequences
• National unity

The reality is that we solved the specific problem that occurred on September 11 rather better than we solved the general problem it exposed. It is now extraordinarily difficult to hijack four airliners simultaneously and use them as guided missiles. Cockpit doors, passenger behavior, intelligence and screening make a literal replay of the events of 9/11 unlikely. The absence of another comparable attack over 25 years shouldn’t simply be waved away as luck.

It still remains the case that a small, determined organization could exploit the enormous complexity and openness of modern American society, while American institutions had great difficulty recognizing a threat that did not fit their existing categories. That problem hasn’t gone anywhere.

American military presence in the Middle East

Before the September 11, 2001 attacks the U. S. is estimated to have had around 25,000 troops in the Middle East; now it has double that. Our presence in the Middle East is now more greatly institutionalized and dispersed.

Islamist terrorism

After 2001 roughly a quarter million people have been killed in Islamist terrorist attacks.

Intelligence

Our intelligence gathering has been reorganized greatly over the last 25 years.

Airport security

Airport security and screening is much more visible and intrusive than it was 25 years ago. Although around $4 billion is collected per year in federal security fees associated with each departure, we spend an additional $8 billion per year on airport security and screening.

Surveillance

One of the greatest changes on the domestic scene over the last 25 years is in the area of surveillance of people in the U. S. Authorities and institutional arrangements created or expanded after 9/11 have become normal parts of American government while technology has made surveillance vastly easier.

Federal bureaucracy

The creation of the Department of Homeland Security following 9/11 consolidated 22 federal agencies and departments. GAO has documented persistent problems integrating their acquisition, financial, information technology, and personnel systems. That is broadly consistent with what I predicted at the time: consolidation would add another bureaucratic layer rather than actually simplify the federal government’s security apparatus.

War powers

The post-9/11 period greatly normalized presidential military action without anything resembling the notional congressional declaration-of-war process.

Fiscal consequences

Federal post-9/11 war expenditures and obligations have been about $8 trillion, including military operations, homeland security, veterans’ care and other costs.

National unity

The increased national unity in the immediate aftermath of the attacks was fleeting. The lamented blog American Future documented that the New York Times started running editorials against George W. Bush about six weeks after the attacks.

I think a lot of the actions of the TSA are more theater than producing security. For two years I unknowingly carried an X-Acto knife in my computer bag while flying. I didn’t discover it because airport security found it. I discovered it myself. During those two years the bag had repeatedly gone through airport screening. If a security system designed in response to 9/11 could repeatedly miss a boxcutter carried by someone who wasn’t even trying to conceal it, exactly what did all of that screening accomplish?

We changed our institutions much more than we changed ourselves.

When I read Francis Fukuyama’s op-ed in the Washington Post advocating what he calls a “realism of means” one sentence took me aback:

The United States has always seen the promotion of liberal democracy as part of its national identity.

because it is objectively untrue. Since the Founding Americans have regarded our own political system as liberal and republican/democratic and have often regarded that system as an example for other peoples. Actively promoting liberal democracy abroad as an objective of American foreign policy has largely been a feature of post-World War II American views. Woodrow Wilson advocated it but was unable to establish his conception of America’s international role as a durable national consensus. John Quincy Adams’s 1821 formulation that America “goes not abroad, in search of monsters to destroy” captures the distinction almost perfectly.

There is another weakness in Dr. Fukuyama’s “realism of means”: he proposes no limiting principle. He provides no mechanism for preventing his “realism of means” coupled with idealism of ends resulting in overextending our means to accomplish those ends. Dr. Fukuyama mistakes one American political tradition for the American national identity. More importantly, his “realism of means” contains no principle for deciding when desirable ends cease to be American interests. Realism can certainly tell us that our means are inadequate to an end. But unless it is permitted to constrain our ends as well, it cannot prevent the accumulation of objectives that eventually exceeds our means. John Quincy Adams understood that problem two centuries ago: America could wish liberty well everywhere without assuming responsibility for securing it everywhere.

I also wanted to comment on Karl Rove’s most recent Wall Street Journal op-ed. In the op-ed Mr. Rove assesses the state of the midterm elections as they enter their final phase. Mr. Rove opens by characterizing the Democratic candidates’ overall strategy:

Democrats appear to believe it’s sufficient simply to blame President Trump for the economy and offer hazy promises to do something about housing, grocery, healthcare and gasoline prices. No need to explain what Democrats would do, how it would work, or what it would cost. Just trash Mr. Trump and chant “affordability.”

citing the rather small advantage Democrats seem to be showing in the opinion polls, and continuing with the Republicans’ overall strategy:

Still, the Republican campaign strategy doesn’t look much better. It’s good news that Mr. Trump finally announced Friday that he’ll open his record $1 billion campaign war chest. He says he’ll spend up to $500 million from his political-action committees now and keep the balance, perhaps for the 2028 election. But it’s already late and could prove too little.

This close to Election Day, advertising airtime sells at a premium. While candidates and party committees by law can buy ads at the lowest unit rate a TV station charges for any nonpolitical client, a PAC has to buy what the market will bear. If Mr. Trump’s MAGA Inc. plans to spend $10 million on the Texas Senate race across the next two weeks—as its Federal Election Commission filings indicate—Mr. Talarico can probably match it by spending $4 million to $6 million.

It all reminds me of the late Mayor Daley’s wisecrack, “No matter how it looks now someone will win the election.”

I have no idea whether the Democrats will take control of the House or the Republicans will retain it. The Polymarket and Kalshi prediction markets presently favor Democratic control of the House, but the actual shift in seats they imply is comparatively modest. That would hardly be an unusual result for a midterm election.