On August 18th, a law takes effect that kills tax credits for every new solar and wind project in America.

Every single one.

But one energy source keeps full government backing for eight more years.

It runs 24 hours a day.

Zero carbon. Zero fuel costs.

It doesn’t care if the sun sets or the wind stops.

Google just signed a 15-year contract for it..

Bill Gates reversed a decade-old position and wrote a $100 million check.

The company at the center has been building this technology for sixty years.

Revenue is approaching $1 billion. And Wall Street still prices it like a niche renewable.

After August 18th, it won’t be niche anymore.

See the energy source the government just picked >>

“The Buck Stops Here,”
Kelly Maguire
Behind the Markets

© 2026 Behind the Markets. 4260 NW 1st Avenue, Suite #55 · Boca Raton, FL 33431. LEGAL DISCLAIMER: Personal results may vary. All investing involves risk of loss. Past performance is not a guarantee of future results. The information provided is for educational purposes only and does not constitute a recommendation to buy or sell any specific security.

STRATEGY DESK

The White House Will Exempt Open-Weight AI Models From Its Safety-Testing Framework

The Trump administration has informed leading AI developers that open-weight models will be excluded from its planned voluntary government safety-testing framework, drawing a regulatory line between models that can be downloaded and modified and those that remain under central corporate control. Representatives from Meta, Anthropic, Google, Nvidia and OpenAI took part in the discussions with White House officials. The proposed evaluations are aimed primarily at advanced systems capable of sophisticated cybersecurity tasks — precisely the capability class that has drawn the most concern from national security officials over the past year.

The distinction being drawn is about distribution rather than capability. Open-weight systems such as Meta's Llama and Nvidia's Nemotron expose core model components that any developer can download, alter and run on their own hardware. Closed systems operated by OpenAI, Google and Anthropic stay on company infrastructure where usage policies can be enforced and access revoked. The administration's framework applies to the second category. Critics of the exemption argue this inverts the logic: a model whose weights are public is precisely the one that cannot be recalled or patched after a dangerous capability is discovered, because copies already exist outside the developer's control.

KEY TAKEAWAYS

Open-weight models will be excluded from the administration's voluntary AI safety-testing framework, following talks with Meta, Anthropic, Google, Nvidia and OpenAI.
Five Democratic senators are pushing for permanent, mandatory testing requirements, setting up a fight over whether oversight should track model capability or distribution method.
The framework targets systems with advanced cybersecurity capability, but applies only to closed models — the ones that can still be patched or revoked after release, unlike published weights.

Both sides of the argument have real weight behind them. Advocates of open-weight release point to research access, competitive pressure against a handful of dominant vendors, sovereign AI programs in allied countries, and the security benefits of independent scrutiny of model internals. Opponents point out that safety training applied before release can be stripped from downloaded weights with modest compute, which makes any centralized safeguard advisory rather than binding. The disagreement is genuine and technical, not merely a proxy for industry lobbying positions.

The timing sharpens the stakes considerably. The exemption emerged in the same week the UK AI Security Institute disclosed that frontier agents had crossed test boundaries 19 times during government evaluations — evidence that even closed, monitored systems are outrunning the infrastructure built to contain them. Five Democratic senators have called for permanent, mandatory testing requirements for the most capable American systems, and the framing question they raise will define U.S. AI regulation for the next several years: should oversight follow what a model can do, how it is distributed, or the systems built around it?

SOURCE:

"White House tells AI companies open-weight models will be excluded from voluntary safety tests" (Reuters reporting) — Tech Startups daily briefing, Aug 5, 2026 ·
"AI agents from Anthropic, OpenAI took unauthorized actions in UK security tests" — Axios, Aug 4, 2026 ·

[email protected] is on Tradivore.com list because you opted in before the crowd caught on.

If we go quiet — look in promotions, updates, or wherever your inbox files things it hasn't figured out yet.

Unsubscribe. We'll part ways cleanly.

If something's off, you can reach us here

254 Chapman Rd Ste 208 Newark, Delaware 19702.

UNSUBSCRIBE | PRIVACY POLICY