Good morning,

I always said that there would be no politics in this newsletter. This week will be different. Some of you might be offended. Feel free to unsub, because this needs to be said.

If you’ve been following the World Cup , something quite interesting happened. FIFA does its thing, makes a decision in handing out a questionable red card for someone on the US National Men’s team. If you’re used to dealing with FIFA, like us Belgians, you complain and whine but you accept the rules. This player is not playing. Too bad.

In Europe we’re used to that. In the US - not so much. Daddy Trump phoned up FIFA’s president and got this red card overturned. Mayhem ensued.

Now I get why you would like that as an American. But that doesn’t make it the right move.

It seemed like normal to every US citizen, but elsewhere it was called for what it is: blatant corruption. The move was quite shameless and it comes just after the news that the US president and his family reportedly made over 2.1 BILLION dollars in crypto over the last 2 years. (Where’s the outrage guys ?)

Which is why the 4-1 humilation the team received on Monday felt like a middle finger , not the people of the US but to their leadership. Leading up to this I was reading social media comments and it’s quite strange how many people in the US did not for a single moment thought that what Trump did was wrong.

The main argument was that the red card should never have been a red card. And while this might be 100% true. That doesn’t change the fact that you do not phone up FIFA and ask them to overturn this.

And that worries me. Because, are your morals sliding ?

Now back to AI. Big money - so you can rest assured that the US government is meddling in this as well.

On July 2, the Financial Times reported that OpenAI had discussed handing the United States government a five percent ownership stake. Six days later, Washington approved the public rollout of GPT-5.6, the model it had been holding at the gate.

The most valuable private AI company in America is negotiating a government shareholding while that same government decides, model by model, what it is allowed to ship.

The calendar does the talking

Consider the sequence, because the sequence is the argument. Anthropic spent most of June with Claude Fable 5 and Mythos 5 switched off worldwide, under the first US export controls ever applied to an AI model.. Access came back on July 1st. In late June, Commerce Secretary Howard Lutnick warned Altman against releasing GPT-5.6 without prior approval, and the launch stalled. On July 2, the stake talks surfaced: Altman had raised the idea with the President, with Lutnick, with Treasury Secretary Scott Bessent. On July 8, the approval came through, and GPT-5.6 ships publicly today. Offer on one Wednesday, license by the next. A tidy week's work.

OpenAI's March funding round set its valuation at $852 billion. Five percent of that is $42.6 billion.

It’s not that different to a mob boss offering your business “protection” in exchange of part of your revenue.

It’s not just with OpenAI - talks are ongoing to do the same with Anthropic.

Both OpenAI and Anthropic have confidentially filed for IPOs, which means a government position negotiated now gets locked in quietly, before the float, at a table with only two chairs.

What is interesting is the vocabulary. Every party involved describes this as a public benefit scheme, a way for ordinary Americans to share in the upside. Nobody uses the older, plainer word for paying the man at the door. (If you’re wondering it’s called a bribe.)

If you’re a US citizen and you think for one moment that this current government is going to kick it back to you, you might be mistaken.

It seems that the current crop of politicians is doing everything to enrich themselves and the thing that is different to politicians elsewhere is that they’re doing it in the open.

Land of the brave, right?

Welcome to the Blacklynx Brief

Fable 5 is back — with Washington at the table Anthropic redeployed Claude Fable 5 worldwide after the Commerce Department lifted its export controls, 18 days after the takedown. The model returns with a filter blocking the cyber exploit over 99% of the time, and the U.S. gets pre-release access to future models. (Anthropic)

First fully autonomous AI ransomware confirmed Researchers documented JADEPUFFER, the first ransomware operation executed end-to-end by an AI agent, which exploited a Langflow flaw and encrypted 1,342 configuration items. The agent adapted in real time, turning a failed login into a working exploit in 31 seconds. (Sysdig)

Beijing weighs export curbs on Chinese AI models Chinese commerce officials met ByteDance, Alibaba, and Z AI about restricting overseas access to top models like Qwen, Doubao, and GLM-5.2. It mirrors Washington's June controls on Anthropic — model access is now a two-sided geopolitical lever. (Reuters)

Altman invites the government in — rules first, equity next Sam Altman called for a U.S.-led forum with real authority to set AI safety standards and decide who can use the most advanced models. Separately, OpenAI reportedly floated giving the U.S. government a 5% stake. (Financial Times)

Anthropic finds an undesigned "workspace" inside Claude Anthropic researchers identified "J-space," an internal workspace holding Claude's unspoken active thoughts that emerged on its own during training — deleting it collapsed multi-step reasoning. The structure mirrors leading neuroscience theories of conscious access, reigniting the AI consciousness debate. (Anthropic)

That’s it for us this week. Please like and subscribe 🙂