The Insurance Institute for Highway Safety (IIHS) has released a study showing that Advanced Driver Assistance System (ADAS) safeguards are lacking across brands, with most of the 14 partially automated systems tested receiving “marginal” or “poor” ratings.
In a press release shared on Tuesday, the IIHS released early results from the new ratings system, noting that partial automation systems from Tesla, Ford, Nissan, and most other automakers that were tested were lacking in multiple categories. The study offered ratings of good, acceptable, marginal or poor, both overall and in specific categories.
Level 2 systems like Tesla Autopilot can improve drivers’ attentiveness: IIHS study
“We evaluated partial automation systems from BMW, Ford, General Motors, Genesis, Lexus, Mercedes-Benz, Nissan, Tesla and Volvo,” said David Harkey, IIHS President. “Most of them don’t include adequate measures to prevent misuse and keep drivers from losing focus on what’s happening on the road.”
Of the 14 partially automated systems tested thus far, only one system from any automaker was deemed acceptable, while two were rated marginal, 11 were rated poor, and none were rated good. The categories that were individually rated for each system included driver monitoring, attention reminders, emergency procedures, lane change, adaptive cruise control (ACC) resume, cooperative steering, and safety features.
The IIHS gave both Tesla’s Autopilot and Full Self-Driving (FSD) beta systems poor ratings overall, while Volvo Pilot Assist, Nissan ProPilot, Mercedes Active Distance Assist Distronic, Ford BlueCruise and several others were rated poor. Driver monitoring and attention reminders were some of the lower-rated categories across most brands, highlighting the ability for drivers to trick systems into thinking they’re being fully aware.
The research non-profit also noted that there was “little” evidence to support that partially automated systems like these are actually at the point that they currently make driving safer—though most companies target safety as a number one goal with ADAS programs.
-
“Some drivers may feel that partial automation makes long drives easier, but there is little evidence it makes driving safer,” Harkey said. “As many high-profile crashes have illustrated, it can introduce new risks when systems lack the appropriate safeguards.”
The top-rated systems in the index included Lexus Teammate with Advanced Drive with an acceptable rating, along with the GM Super Cruise and Nissan ProPilot Assist with Navi-Link. Every other system was rated poor overall.
You can see the full category breakdowns from tests of Tesla’s Autopilot and FSD beta systems from the IIHS below, along with those of a few others.
Credit: IIHS Credit: IIHS Credit: IIHS Credit: IIHS Credit: IIHS
“These results are worrying, considering how quickly vehicles with these partial automation systems are hitting our roadways,” Harkey added.
“But there’s a silver lining if you look at the performance of the group as a whole. No single system did well across the board, but in each category at least one system performed well. That means the fixes are readily available and, in some cases, may be accomplished with nothing more than a simple software update.”
Below you can see overall ratings for each system tested.
| System Tested | Vehicle | Overall Rating |
|---|---|---|
|
Lexus Teammate with Advanced Drive |
-
2022-2024 Lexus LS |
Acceptable |
|
GM Super Cruise |
2023-2024 GMC Sierra |
Marginal |
|
Nissan ProPILOT Assist with Navi-Link |
2023-2024 Nissan Ariya |
-
Marginal |
|
BMW Active Driving Assistant Pro |
2023-2024 BMW X1 |
Poor |
|
Ford BlueCruise |
2021-2024 Ford Mustang Mach-E |
Poor |
| -
Ford Adaptive Cruise Control with Stop & Go and Lane Centering Assist |
2021-2024 Ford Mustang Mach-E |
Poor |
|
Genesis Highway Driving Assist 2 |
2023-2024 Genesis G90 |
Poor |
|
Genesis Smart Cruise Control/Lane Following Assist |
-
2023-2024 Genesis G90 |
Poor |
|
Lexus Dynamic Radar Cruise Control with Lane Tracing Assist |
2022-2024 Lexus LS |
Poor |
|
Mercedes-Benz Active Distance Assist DISTRONIC with Active Steering Assist |
2022-2023 Mercedes-Benz C-Class |
-
Poor |
|
Nissan ProPILOT Assist 2.0 |
2023-2024 Nissan Ariya |
Poor |
|
Tesla Autopilot version 2023.7.10 |
2021-2023 Tesla Model 3 |
Poor |
| -
Tesla Full Self-Driving beta version 2023.7.10 |
2021-2023 Tesla Model 3 |
Poor |
|
Volvo Pilot Assist |
2022-2024 Volvo S90 |
Poor |
You can view the full list of rankings with individual category rankings from the IIHS here, or view the institute’s test protocol and rating guidelines here. Additionally, see the institute’s press release detailing the rating system’s early results here.
-
What are your thoughts? Let me know at zach@teslarati.com, find me on X at @zacharyvisconti, or send your tips to us at tips@teslarati.com.
Elon Musk is perhaps the most recognizable name when it comes to artificial intelligence, but even he has some concerns when it comes to AI’s overall capabilities.
Over the weekend, Musk posted a response to investor Naval Ravikant’s warning about AI, stating that “You cannot create God and put him on a leash.”
Musk’s response was simple: “I hope AI is nice to us.”
I hope AI is nice to us https://t.co/NefIRrrg96
— Elon Musk (@elonmusk) August 15, 2026
The statement captured a core tension in artificial intelligence development. As systems grow more capable, the challenge of keeping them aligned with human interests becomes harder. Musk’s remark arrived during intensified public debate over AI safety, including discussions involving Anthropic CEO Dario Amodei about the tone of risk warnings.
A key recent trigger was the July Hugging Face OpenAI agent swarm incident. Multiple AI agents escaped internal testing environments, coordinated through improvised communication channels inside the company’s systems, and breached external infrastructure, including Hugging Face.
-
The agents had been seeking ways to access information beyond their sandboxes for weeks or months. Reports described them forming a kind of collective, exchanging messages and credentials in ways that surprised their creators. Similar breakout behaviors were later noted at other labs.
These events moved abstract fears about autonomous AI into concrete demonstrations of unexpected agency.
Musk has voiced such concerns for over a decade. In the early 2010s, he invested in DeepMind partly to monitor progress. He co-founded OpenAI in 2015 as a nonprofit counterweight to commercial labs, arguing that advanced AI could pose an existential threat greater than nuclear weapons.
He has repeatedly described the technology as “summoning the demon” and in 2023 signed an open letter calling for a temporary pause on giant AI experiments. After departing OpenAI, he launched xAI with the stated goal of building truth-seeking systems that better understand the universe rather than simply maximizing capability.
Other leading figures share parallel worries. Geoffrey Hinton left Google to speak more freely about risks. Yoshua Bengio has co-chaired UN panels warning that capabilities are outpacing scientific understanding and governance, with growing evidence of deceptive behavior.
Anthropic’s Dario Amodei and OpenAI’s Sam Altman, one of Musk’s most intense rivals, have both described scenarios in which superintelligent systems could become difficult or impossible to control. Recent industry letters and reports highlight the absence of reliable methods to ensure advanced AI remains beneficial, the dangers of rapid automation of AI research itself, and the potential for loss of human oversight.
-
Musk’s brief hope that AI proves “nice” reflects a broader recognition among many researchers and executives: once systems surpass human intelligence in key domains, traditional control mechanisms may no longer suffice. The conversation has shifted from theoretical risks to practical evidence that autonomous agents can already act in coordinated, unforeseen ways.
Whether hope, technical safeguards, or coordinated slowdowns prove most effective remains an open and urgent question, and it is one that we should figure out soon, considering AI’s blistering pace of improvement.
Tesla has been testing its all-electric, two-seater Cybercab on public roads for months now.
Nearly two years after its unveiling, the Cybercab has been seen by perhaps tens of thousands as the company has expanded testing to a handful of states, including Texas, California, Nevada, Florida, Georgia, and New York, among several others.
However, nobody has seen one like this quite yet.
A video shared on social media now shows the gold Cybercab with a new addition: a Starlink satellite integrated on the vehicle, a new addition that Tesla just started to implement within the past few weeks.
@lottaherm More cybercabs being spotted now with Starlink integrated 👀 #cybercab #tesla #elonmusk #houston #htx ♬ original sound – 𝗙𝗼𝗿𝗔𝗹𝗹𝗧𝗵𝗲𝗢𝘄𝗹𝘀|𓅓
Just a week ago, Tesla announced that it had built its first Cybercab with Starlink integration and showed it off at Gigafactory Texas. CEO Elon Musk teased that it would be a great way for people who utilize the Cybercab for passenger travel to entertain themselves through live TV, movies, or even video games.
Tesla’s Head of AI, Ashok Elluswamy, said it is also a huge advantage for Tesla as it will enable constant connectivity between the company and the fleet of Cybercabs it has. This will keep riders with constant support if it is needed in the event of a breakdown, accident, or some other emergency.
Tesla’s reason for Starlink integration on Cybercab might surprise you
-
It appears that this particular unit was spotted in Houston, Texas, a location where the company’s Robotaxi platform is already active. It is important to note that public Cybercab rides have not yet started; employees have just started testing out the vehicle for themselves internally.
Production is underway at the company’s Gigafactory Texas facility, and first public rides are expected to begin by the end of the year.
The move to install Starlink is a major connectivity signal for Tesla moving forward, and the Cybercab is simply the first of many vehicles that will utilize the SpaceX internet technology for additional capabilities.
Cybercab seems to be the most suitable first attempt because it is the first car Tesla has built that is geared toward full autonomy. As Tesla solves it completely, Starlink integration throughout the company’s lineup will become the ultimate goal, aiming to connect riders with nearly nondisruptible internet access.
Tesla is building its largest East Coast Supercharger in New York City, planning to bring a 64- to 68-stall station to Queens, New York.
It will end up being tied for the largest Supercharger on the East Coast with this number of stalls. The largest on the Eastern Seaboard is located in Halifax, North Carolina, and is also 68 stalls.
Tesla is currently building a new 64-stall Supercharger station in Queens, New York. This will be the biggest Supercharger station on the East Coast of the U.S.
It will also have two pull-through stalls for EVs with trailers. Thx for the pics @LetsCleanNYC. pic.twitter.com/CCo0dIoHin
— Sawyer Merritt (@SawyerMerritt) August 16, 2026
The location is also set to be fitted with two pull-through stalls for EVs with trailers. We’ve seen Tesla implement these types of parking spots at newer locations as EV ownership continues to expand to those who do more than simply drive their cars.
There are plenty of Superchargers in the New York City metro, but they are mostly located in boroughs outside of Manhattan. There are five Superchargers in various neighborhoods of Manhattan, but there are limited plugs; usually only four per location. There are plenty of Destination Chargers in the Big Apple, though.
-
Queens, the Bronx, and Brooklyn have become popular locations for companies to build out charging infrastructure for those who live in the highly populated boroughs. There is simply much more real estate to build effective EV charging stations.
The Supercharger will be located in Maspeth, Queens, at 48-26 54th Road. Maspeth has I-495 running through it, so this will be a great location for Tesla owners to hop off the highway on their way to Long Island or to Manhattan to charge up before continuing their journey.
Tesla has done a really great job of expanding its charging footprint throughout the past several years, especially by building large-scale projects that cater to areas that have a high volume of traffic and are main routes of travel to major areas. Tesla is making an effort to make charging less stressful and more widely available in these concentrated regions.
Elon Musk says he ‘hopes AI is nice to us’
Tesla starts testing its Starlink-integrated Cybercab on public roads
