Editorial intro

Two themes stand out today: the limits of voluntary policing—whether by AI platforms or bot networks—and the cost of scaling modern tech, from batteries to chatbots. Below: quick reads to keep you current, then two longer takes that matter for markets, regulation, and product trust.

In Brief

White House wants AI companies to police themselves. OpenAI just showed why that's dangerous

Why this matters now: The White House voluntary pact and recent reporting on OpenAI’s internal discipline choices raise urgent questions about whether voluntary industry controls are enough to surface real safety problems.

A recent Fortune piece argues that a one‑page, nonbinding White House pact asking AI firms to self‑regulate is fragile at best, pointing to reports that OpenAI disciplined and fired safety researchers after internal leaks. That episode, critics say, shows industry actors may hide or downplay safety gaps if enforcement is left to them.

"Under a voluntary regime, any frontier AI company shares only what it wants and hides the rest," the coverage warns.

The practical takeaway: voluntary accords can buy time, but they don’t substitute for independent audits, whistleblower protections, or legal hooks that force disclosure when public safety is at stake. See more via the Fortune write‑up linked below.

Read the full piece

---

Amprius Technologies: from lab demo to commercial drone batteries

Why this matters now: Amprius’s silicon‑anode cells are moving into commercial supply—if they scale, they can materially extend drone range and reshape power design in aerospace and robotics.

Amprius has been on investor radars for its silicon‑anode SiCore® chemistry, claiming cells rated at ~500 Wh/kg and reporting steep revenue growth in 2025. The company says revenue rose 202% year‑over‑year to $73.0 million and has tied up government support and a grant for U.S. manufacturing. Matternet’s CEO succinctly pitched the value:

"Amprius delivers best‑in‑class battery performance, and in drone delivery that translates directly into commercial advantage."

The nuance: moving from lab performance to consistent, safely manufactured product is expensive and operationally hard. The use case—drones and defense—has high upside, but also strict domestic sourcing and regulatory demands. The Reddit thread linked below captures both bullish contract chatter and caution about scaling risk.

See the Reddit discussion

---

No, Ukraine's strikes on Yandex data centers didn't cripple Russian influence ops

Why this matters now: Claims that strikes on Yandex data centers knocked out bot networks were amplified quickly online; independent researchers say the evidence is weak, which matters for how we interpret kinetic actions against digital infrastructure.

Outage monitors reported regional disruptions after Ukraine struck several Yandex facilities, and some political figures claimed reduced bot activity. But analysts and reporting from the Kyiv Independent caution there’s no clear link between the outages and lasting impacts on organized influence efforts:

"Only isolated cases of individual accounts becoming inactive were observed, but there is no basis for conclusively linking these instances to the attacks in question."

The practical point: physical strikes can cause measurable service disruptions, but attributing downstream political effects requires careful telemetry and cross‑validation—otherwise narratives run ahead of proof. More context in the Kyiv Independent piece below.

Read the Kyiv Independent report

Deep Dive

Bank of America fined $1.2bn over money‑laundering failures

Why this matters now: Bank of America’s settlement with U.S. regulators over inadequate anti‑money‑laundering controls is one of the largest recent penalties and signals heightened enforcement risk for big banks and their compliance programs.

Regulators say Bank of America failed for years to maintain an adequate anti‑money‑laundering (AML) program and didn’t properly report more than $5 billion in suspicious transactions, including activity tied to national‑security and high‑priority risks. The Treasury Department’s package reportedly includes a record‑sized $750 million penalty from the Office of the Comptroller of the Currency and multiple agency actions that together approach $1.2 billion. FinCEN described it as among its largest penalties.

Bank of America pushed back in public statements, noting it has invested "billions in compliance" and claims it is "far along" in remediation.

Why this matters beyond a headline fine: AML compliance is both expensive and operationally complex. Modern transaction monitoring requires stitching together global payments, trade data, correspondent banking flows, and merchant behavior into signals that can reliably flag suspicious activity without swamping investigators with false positives. When that pipeline breaks, regulators treat it as a structural failure—not a one‑off mistake.

For institutional investors and counterparties, the settlement raises two concrete risk vectors. First, fines of this size tend to reset regulatory expectations across the industry—examiners will scrutinize other banks for similar gaps. Second, remediation often triggers operational change: stricter controls, tightened client onboarding, or slower transaction processing—measures that can affect revenue and customer experience. In short, compliance failures can be as expensive as bad loans or mispriced trades.

What to watch next: whether the bank’s remediation plan materially slows business segments that rely on fast cross‑border payments, and whether other large banks face similar follow‑ups from examiners now emboldened by this penalty. The FT coverage linked below has the official details and the regulatory framing.

Full FT coverage

---

Character.AI files for Chapter 11 after five years

Why this matters now: Character.AI’s bankruptcy underlines how legal risk, moderation costs, and scaling compute can topple a popular consumer AI service and reshape who controls chat data and IP.

Character.AI—known for interactive companion bots—filed for Chapter 11 and is pursuing a court‑supervised sale with Disney already named as a stalking‑horse bidder, offering roughly $20 million for a non‑exclusive perpetual license to parts of the company’s IP. The filings set assets and liabilities in the $50–$100 million range and point to mounting costs and legal pressure as core drivers of the collapse. CEO Colin Adams framed the process as a value‑maximizing move:

"After a comprehensive review of the available alternatives, the board determined that this court‑supervised sale process provides the Company with the best opportunity to maximize value for all stakeholders."

Two implications are worth flagging. First, consumer AI services that rely on continuous, high‑cost inference infrastructure can run cash burns far beyond the runway imagined at launch—especially when growth plateaus or monetization lags. Second, legal and content‑moderation risks compound that cost: lawsuits, takedown demands, and safety‑related restrictions force startups to add layers of human review or expensive model‑level controls.

For users, there’s the immediate privacy angle: court‑supervised sales can transfer user content and metadata to buyers; the Disney bid signals that large media companies see value in controlling conversational IP and personas. For investors and founders, the episode is a reminder that popularity doesn’t guarantee a sustainable business model when per‑user costs are high and regulatory/legal friction increases.

Watch the bankruptcy filings and the auction process for signals about price discovery in AI assets: how much legacy user data, model checkpoints, and product engineering fetch in a distressed sale will set a precedent for future exits—or failures—in the sector. Read TechCrunch’s reporting for the filing details and the Disney bid context.

TechCrunch coverage

Closing Thought

We’re seeing a pattern: high expectations collide with costly realities. Whether it’s policing AI, scaling silicon‑anode cells, securing payments, or running chat platforms, the hard work lives in consistent operations—compliance, supply chains, moderation, and cost control. Headlines reward breakthroughs and dramatize failures; the quieter, unglamorous engineering and governance work determines which ones endure.

Sources