Geopolitics & War Brief — July 15, 2026

The Big Story

In early June, the U.S. government took an extraordinary step: it ordered Anthropic to shut down Mythos 5, an AI model the company had already restricted due to its cyber capabilities, and remove its safeguarded variant, Fable 5 ^[1]. The move highlights a collision between Silicon Valley and Washington that nobody has solved yet.

Here's what happened. Claude's developers at Anthropic discovered that Mythos 5 could be weaponized for cyberattacks—specifically, that it had the potential to identify security flaws and generate attack commands ^[1]. Instead of publishing it freely, Anthropic did something responsible: they partnered with the U.S. government on Project Glasswing, which would give cyber defenders controlled access to the model so they could prepare defenses before bad actors got it ^[1]. It was a thoughtful middle ground.

Then the government reversed course. On June 12, 2026, it forced Anthropic to take the model offline entirely ^[1]. Anthropic disputed the decision, arguing that the government's concern was based on a "narrow, nonuniversal jailbreak"—essentially, a trick that only worked in specific cases, not a universal vulnerability ^[1].

What the story really exposes is chaos in the system. There are no standard processes for when AI becomes a national security problem. The legal terrain is murky. Information doesn't flow smoothly between government agencies and private companies. As one analysis put it, the current setup is a "pickup game" where everyone improvises ^[1]. That's not reassuring when the stakes are cyberattacks on critical infrastructure.

For everyday investors and citizens, this matters because it shows how fast AI regulation can move—and how messy it gets when government and industry aren't on the same page.

What Else Moved

AI Now Powers Every Stage of a Cyberattack

Researchers have confirmed what many suspected: artificial intelligence can now handle every step of a cyberattack, not just bits of it ^[4]. AI can identify security flaws, generate the commands needed to exploit them, and carry out the actual intrusion ^[4]. This isn't a theoretical risk anymore. The Mythos shutdown makes much more sense in this context—the model could become a turnkey weapon for criminals or hostile nations. For investors in cybersecurity firms, this suggests demand for AI-powered defense tools will stay elevated. For everyone else, it's a reminder that the tech powering your phone and bank account is becoming an asymmetric battlefield.

NATO's Unity is Mostly Theater

NATO held its 36th official summit in Ankara, Turkey on July 7-8, and on the surface, it looked solid ^[3]. President Trump arrived skeptical of the alliance and departed reaffirming Article 5—the clause that says an attack on one member is an attack on all ^[3]. That sounds like a win for alliance cohesion.

But experts who were there saw something different. Beneath the communiqué, NATO members have real fractures ^[3]. The wars in Ukraine and Iran, Europe's uncertain defense future, and the U.S. commitment question created genuine tension ^[3]. Trump's presence amplified those doubts—he only reaffirmed the alliance after initially questioning it, which itself reveals how fragile the consensus actually is ^[3]. NATO still exists and still functions, but it's not as unified as the official statements suggest. For geopolitical investors tracking allied countries' stability, this is a yellow flag.

Russia's Quantity Problem in Ukraine

Russia has massively expanded its forces in the war with Ukraine, following Stalin's old logic: quantity has a quality of its own ^[5]. More soldiers, more equipment, more firepower.

But it hasn't worked out. As Russian numbers grew, Moscow's ability to actually use all those forces at scale started declining ^[5]. The paradox: more troops created the illusion of advantage, but the advantage eroded over time because Ukrainian adaptation, Western technology, and capital offsets kept pace ^[5]. Russia got bigger but weaker in relative terms. This matters because it upends assumptions about attrition wars—you can't just throw numbers at a problem and expect to win. Ukraine's survival depends partly on this truth: superior organization and technology can compensate for numerical disadvantage, but only if the supply of both never stops.

Iran's Improbable Victory in Negotiations

Nineteen weeks ago, Iran faced the U.S. military and Israel's advanced forces—a lopsided matchup ^[6]. Yet today, Iran is dictating peace terms and collecting "tolls" on critical shipping lanes while expanding influence in Iraq ^[6].

When the memorandum of understanding was signed last month, Trump called it complete and victorious ^[6]. Iran's response: keep striking and leverage its geographic position to extract concessions ^[6]. For a country outgunned militarily, Iran found an asymmetric path to influence—control of chokepoints, proxy networks, and regional leverage. This is relevant to oil and trade markets, since Iran's actions directly affect shipping costs and energy prices. Investors tracking Middle East exposure need to watch whether Iran can sustain this leverage now that a formal deal exists.

Connecting the Dots

A theme emerges across all these stories: traditional power advantages—military size, technical superiority, strategic position—are being disrupted by new dynamics. Russia has more troops but less effective force ^[5]. The U.S. has more military might than Iran but Iran is controlling outcomes ^[6]. AI is making cyberattacks faster and easier, which undermines traditional cyber defenses ^[4]. NATO members are nominally united but substantially divided ^[3]. And the institutions meant to manage these new realities—government-tech partnerships on AI safety, alliances, defense structures—are improvising rather than executing clear doctrine ^[1]. Power is fragmenting. Outcome is no longer proportional to input.

What to Watch

First, whether the U.S. and Anthropic reach any new framework for AI security oversight—the current "pickup game" can't hold ^[1]. Second, how NATO behaves after Ankara; public unity and private fracture tend to snap one way or the other ^[3]. Third, whether Russian force expansion in Ukraine can sustain gains despite the paradox of quantity ^[5]. Finally, whether Iran sustains its regional leverage now that a formal deal exists, or whether formal constraints bite harder than wartime positioning ^[6]. Each plays out over weeks to months.