Half of Americans think so, according to a recent survey by Talker Research. (And a quarter of Americans think absolutely otherwise. Yowza.)

My favorite stat among the findings? That 4 in 10 Americans believe that past behavior to AI agents will one day be taken into account.

Scene: Saint Peter at the pearly gates. “Denied because you were a bit huffy with Siri when she couldn’t dig up the right phone number. To the bad place.”  —Andrew Nusca

P.S. A few longtime readers wrote in to lament our breezier format, saying they miss the longer analytical essays. To get your fix, please subscribe to Fortune’s twice-weekly Eye on AI newsletter. I think you’ll like it.

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

NHTSA investigating Tesla self-driving software

You might want to wait a minute to turn FSD back on in your Tesla. After four incidents—including one in which a pedestrian was killed—the National Highway Traffic Safety Administration said it was conducting a preliminary investigation into Tesla’s “full self-driving” software, which has been installed in some 2.4 million vehicles, according to a notice the U.S. regulator issued last week. 

The agency said there were four specific crashes where a Tesla vehicle was using FSD in low-visibility (think foggy or dusty) conditions. One of the crashes resulted in a fatality, and another involved a reported injury, according to the agency, which said it would be evaluating  whether Tesla’s FSD engineering controls were able to detect and respond to these kinds of lower-visibility conditions the way they should.

The timing isn’t great for Tesla. Earlier in October, CEO Elon Musk announced plans to bring so-called unsupervised FSD—as in “Look Ma, no hands!”—to Tesla’s Model 3 and Model Y in Texas and California as early as next year. It’s not the first time NHTSA has investigated Tesla’s driver assistance features, but this latest probe could put a damper on Musk’s plans—at least temporarily. —Jessica Mathews

How iPhone got its groove back

Apple’s iPhone is back, as far as Chinese consumers are concerned.

According to Counterpoint Research, the iPhone 16 has sold 20% more in the country during its first three weeks, compared with sales of the iPhone 15 a year ago. Look up to the pricier Pro and Pro Max models, and the picture is even rosier: 44% up year-on-year.

“Given the smooth production ramp-up, consistent pricing strategy, and the initial wave of upgrades by existing iPhone users, the iPhone 16 series has experienced substantial growth in the Chinese domestic market,” Counterpoint analyst Ivan Lam told Bloomberg.

Not bad at all, especially considering that Apple hasn’t yet secured a local AI partner, which it will need to roll out Apple Intelligence features in China. But take this data with two big grains of salt.

First, these are just estimates. We’ll get a clearer picture when Apple reports quarterly earnings on October 31 (though the results will only include about one week of iPhone 16 sales).

Secondly, Apple rivals like Huawei, Xiaomi, and Oppo are due to release their new smartphones shortly. Let’s see how iPhone sales fare in China then. —David Meyer

Chinese robo-taxi firm Pony.ai files for IPO

The autonomous driving company Pony.ai is ready to go public on Nasdaq.

The eight-year-old Chinese company, which is backed by Japanese auto giant Toyota and was valued at $8.5 billion during its last fundraise in 2022, said in an SEC filing that its revenue in the first half of this year nearly doubled, to $24.7 million. It’s not yet profitable.

Pony.ai runs a fleet of some 250 Toyota- and Lexus-branded robo-taxis. It has offices in Beijing, Guangzhou, and Fremont, Calif. and has been road-testing in them and elsewhere. (Though not without issue: A 2021 crash led to a permit suspension in California.)

The announcement ticks a lot of boxes for the present state of the technology industry, if not the broader economy. AI? Check. Robo-taxis? Check. Unicorn IPO? Check. China? Well…we’ll allow it. 

But the category’s risks remain clear. As the company’s filing states, “our technology targeting Level 4 autonomous vehicles requires significant investment and may never be commercially successful on a large scale, or at all.” 

The proposed ticker, by the way? PONY, of course. —AN

Bluesky capitalizes on rivals’ missteps

The decentralized social media service Bluesky now has more users than ever, reaching 12 million last week. It became the third most downloaded free app in the Apple app store. 

It likely has rival X, and to a somewhat lesser extent Threads, to thank for a surge of more than one million new signups in the span of a few days.

Bluesky has seen such user surges before, like when X was blocked in Brazil for owner Elon Musk’s initial refusal to pay a fine over its failure to moderate content in the country; about 2 million new Bluesky users showed up. Or when Musk decided to impose a “rate limit” on all users—Bluesky saw so much new traffic it crashed briefly.

Now X has decided users don’t really need the ability to completely block accounts, a basic feature of all social media platforms. (Even Donald Trump’s Truth Social lets users fully block any account.) 

Combine that with Threads’ recent failures with content moderation—users are being suspended or having posts deleted for using innocuous words like “cracker” or expletives—and it’s no surprise why people are decamping to Bluesky. 

The company is certainly enjoying every moment. —Kali Hays

Perplexity launches enterprise search

Enterprise search, or tools for searching internal company documents, was a big focus for startups in the Web 1.0 era. Two decades later, it’s buzzy again.

Perplexity, the popular AI-powered search engine, announced last week a push into the category with a tool that lets users search across both public web content and corporate information. The product, called Internal Knowledge Search, is available to paid Perplexity Pro and Enterprise Pro users. 

Companies usually have vast troves of internal documents filled with vital information that are difficult to search. Perplexity’s new tool is aimed at solving that problem by helping workers find what they need within their organizations, combined with relevant external information from the web. 

There is already plenty of competition in the AI enterprise search space, including startups like Glean, Coveo, Elastic, and Guru. But as anyone who’s used an internal search tool at work knows, the problem is far from solved—even after decades of companies trying. —Sharon Goldman

More data

—U.S. Treasury stopped $4b in fraud. Machine learning FTW.

—Samsung postpones equipment deliveries for chip factory. Customers, where art thou?

—Has TSMC been making chips for Huawei? The U.S. investigates.

—Intel trying to sell Altera stake for billions. Minority or majority, they’re easy.

—Verizon will buy $1b in spectrum from US Cellular. It’s contingent on another deal with T-Mobile.

Endstop triggered