Our analysts track the weekly developments reshaping AI and how it reaches users. Brands building AI products can partner with vetted AI companies to bring their ideas to life.
The government loosened its grip on Anthropic's top models.
OpenAI answered with an offer of its own, a slice of the company handed straight to Washington.
Meta, short on AI compute two weeks ago, says it has plenty to sell now.
This week, ownership became the bargaining chip in AI's dealings with Washington.
Here's where things stood in more detail.
OpenAI Floats a 5% Government Stake
OpenAI proposed giving the U.S. government a 5% equity stake, worth roughly $42.6 billion at its $852 billion valuation.
The plan would extend to other major AI labs, each ceding a similar share through a sovereign wealth fund modeled on the Alaska Permanent Fund.
Sam Altman pitched the idea directly to President Donald Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent.
Anthropic is not part of the talks, according to a source close to the matter.
The 5% figure did not come from nowhere.
Bernie Sanders has floated a far bigger idea, a one-time 50% tax on major AI companies' stock funneled into a public wealth fund.
Sanders and Altman met privately for about an hour at his Senate office, and Sanders came away unconvinced.
"The issue is not just, I'm going to give you a couple of thousand dollars a year," Sanders said in a WIRED interview.
"The issue is power, and they want to keep that. Money is one thing. Power is another thing."
The 5% stake only buys favor with the administration that accepts it, and the one after that inherits none of the obligation.
It looks like protection, yet it locks in no policy, no exemption, and no guarantee against the next crackdown.
Fable 5 Is Finally Back
Washington lifted its export controls on Fable 5 and Mythos 5 on June 30, ending a 19-day shutdown, per Anthropic.
Fable 5 is back globally as of July 1, while Mythos 5 stays limited to the roughly 100 organizations approved on June 26.
The trigger was a jailbreak report from Amazon researchers, flagged to officials by Amazon, which is also an Anthropic investor.
Citing national security, the government ordered access cut for every foreign national, so Anthropic pulled both models worldwide to comply.
Anthropic then retrained a safety classifier that now blocks the technique in more than 99% of cases.
"Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America's leadership in AI," Lutnick wrote in an X post.
Enterprise teams took notice.
Forrester's Allie Mellen told Bloomberg that companies are spreading workloads across several models to avoid a repeat.
Nineteen days offline turned one AI vendor into a business risk that companies now have to plan around.
Meta Plans to Sell Its Spare Compute
Two weeks ago, Meta was short on compute. Now, it wants to sell surplus.
Meta is building a cloud business, internally called Meta Compute, to sell excess AI capacity.
The move would pit Meta directly against AWS, Azure, and Google Cloud.
View this post on Instagram
Investors reacted fast, and CoreWeave fell as much as 15% while Meta rose about 9%.
The sell-off hit the neoclouds hardest, since investors see providers like CoreWeave as the ones Meta would undercut.
The surplus is a byproduct of how hard Meta has been building.
The company is on track to spend as much as $145 billion on AI infrastructure this year, up from about $72 billion in 2025.
This much capacity outruns what even Meta's own products can use, which leaves room to rent the rest out.
"It's definitely on the table," Zuckerberg told investors at Meta's May shareholder meeting, addressing whether the company might enter cloud computing.
Bloomberg Intelligence's Mandeep Singh called Meta late to renting out compute.
He said an agentic AI product would have excited investors more than spare capacity.
Selling spare capacity is a hedge, and Meta reaching for it says the surest money in AI right now is still the infrastructure.
Three Ways to De-Risk Your AI Stack
Government ties to AI labs and infrastructure capacity are both moving faster than most vendor strategies account for.
Three moves matter most right now:
- Favor vendors that push for durable rules. A lab chasing one administration's goodwill leaves you exposed the moment power changes hands.
- Write exit ramps into capacity contracts. Supply and pricing can flip in a single quarter, so keep a second provider warm.
- Trace who funds and who audits your vendor. When an investor and a regulator overlap, their interests can collide, and your access pays for it.
The smart move is to assign someone to watch your vendor's government and funding ties year-round, the same way finance tracks a key supplier.
Our Take: Which Approach Really Fixes the Access Problem?
The access problem is bigger than any one company's deal with Washington.
We think neither OpenAI's stake nor Anthropic's standardization push has fixed it yet, and that's what most coverage missed this week.
View this post on Instagram
Giving the government a stake in one company sets no precedent for the next export order, jailbreak report, or vetted-partner list.
Anthropic is pushing the other way, asking for a "durable, transparent process" that would treat every frontier lab the same.
But that process is still a request, not a rule, so it protects no one until Washington actually writes it into policy.
OpenAI is buying a relationship, and Anthropic is asking for a rule, and neither exists in enforceable form today.
Last week's roundup covered Washington's AI vetting lists, Google's Gemini capacity limits, and Gap's multi-vendor marketing stack.
It's the same access thread running through this week.
Does your brand have a plan for what happens when a vendor relationship with the government shapes your AI access?
These leading AI companies help brands move from generic AI tools to specialized models trained on their own data.
