Shashank Joshi

speaker

307 appearances 4 recordings 1 series first heard May 2026 last heard 22 Jul

Shashank Joshi’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 4 in all, peaking in Jul 2026 with 2.

Appearances

newest first · ▶ plays the moment

The Intelligence from The Economist · Roaring trades: oil majors’ secret success story · 2 Jul 2026 podcast

It was pretty carefully restricted, but there was an alleged way to get round some of its safeguards, and that prompted, in a kind of panic, the administration to say, you have to now shut down everything.

We've also seen in recent weeks.

OpenAI, which makes the Chat GPT series of models, release another model called GPT 5.6 or SOOL, a particularly capable variant of that.

And they have not been allowed to freely release it.

Basically, the government has given them permission to give it to us a relatively modest number of trusted partners.

Well, there was an executive order on June 2nd, and it was a way of of trying to impose some sense of governance of AI.

And it said, look, we should have a framework in which private companies can voluntarily share their most advanced models with the government and discuss the risks.

And that was kind of happening anyway, right?

We know that OpenAI was doing that.

We know that Anthropic was doing it.

But the interesting thing about that order is that it explicitly says.

Nothing should be a formal licensing regime.

This is not a compulsory or mandatory system for companies to follow, because there are a lot of people in the Trump administration who basically think this should be unshackled to maximize innovation at a time when investment in AI infrastructure like data centers is really holding up some parts of the American economy.

Effectively, we have a de facto licensing regime in place that is described as

Voluntary, but to me it looks about as voluntary as, you know, paying your taxes.

You can choose not to do it, but there are gonna be some consequences if you don't.

There is a certain sense to it.

You want to exploit this window of superiority in particular ways without giving access to the model to every potential bad actor who would use it for nasty things.

The problem is, though, it's been such an opaque process.

It hasn't been clear what the rules are, and it hasn't been clear who gets special access.

Showing 81–100 of 307 · page 5 of 16