News
Tesla Model 3 gets penalized in Europe despite top scores in vehicle assistance and safety
In collaboration with Thatcham Research, the Euro NCAP has launched the world’s first Assisted Driving Grading system, a new set of metrics that are specifically designed to evaluate the driver-assist systems of cars available on the market today. For its first batch of vehicles, the firms evaluated 10 cars, from premium SUVs like the Mercedes-Benz GLE to affordable hatchbacks like the Renault Clio to all-electric vehicles like the Tesla Model 3.
As noted by Thatcham Research Director of Insurance Research Matthew Avery in a video outlining the results of the Assisted Driving Grading system’s first tests, vehicles would be graded on three metrics: the level of vehicle assistance that they provide, the level of driver engagement that they offer, and the effectiveness of their safety backup systems. The results of these tests, especially on the Tesla Model 3’s part, were rather peculiar, to say the least.
Out of 10 vehicles that were evaluated, the Tesla Model 3 ranked 6th with a “Moderate” grade, falling behind the Mercedes-Benz GLE, BMW 3-Series, and Audi Q8, which were graded as “Very Good,” and the Ford Kuga, which received a “Good” rating. This was despite the Tesla Model 3 receiving the top scores in the “Vehicle Assistance” and “Safety Backup” metrics.
The study, for example, dubbed the Model 3 as outstanding in terms of steering assistance, with the vehicle steering itself exceptionally well through an S-shaped curve at speeds of 80, 100, and 120 km/h. Tesla’s lane change systems were also satisfactory, despite the system’s limitations in Europe. Distance control was dominated by the Model 3 as well, with the evaluators stating that Tesla’s adaptive cruise control featured a “high level of technical maturity.” From a score of 100, Tesla’s vehicle assistance received a score of 87, the highest among the cars tested.
The Model 3’s safety backup systems were also a league above its competition. As noted in a post from the Allgemeiner Deutscher Automobil-Club e.V. (ADAC), Tesla demonstrated its strengths with the Model 3’s collision avoidance systems. The all-electric sedan earned a perfect score in the firms’ tests, outperforming its premium German competition. Overall, the Model 3 received an impressive score of 95 in the Assisted Driving Grading system’s “Safety Backup” metric.
Considering these scores, one might wonder why the Model 3 ended up ranked 6th among the 10 vehicles tested by the Euro NCAP and Thatcham Research. As it turned out, this was because of the Model 3’s poor scores in the “Driver Engagement” metric, where the vehicle only earned a score of 35 out of 100. So poor was the Model 3’s scores in this metric that it was ranked last among the 10 vehicles that were evaluated.
A look at the reasons behind the Model 3’s poor scores in “Driver Engagement” includes a number of interesting insights from Thatcham Research and the Euro NCAP. When testing the vehicles’ steering override functions, for example, the evaluators stated that the Model 3 resisted steering overrides from its driver. These issues were explained in the ADAC’s post.
“Should the driver make a steering movement in order to avoid an object or a pothole in the roadway, the steering assistant should allow this without resistance. In the Tesla Model 3, for example, this is not the case. Apparently, Tesla trusts the system more than its driver. The necessary cooperative assistance is not given. Instead, the Tesla system prevents its driver from attempting to intervene – it mustn’t be,” the ADAC remarked in its post.
Even more interesting is that part of the Model 3’s poor “Driver Engagement” scores was due to the term “Autopilot,” which Tesla uses to describe its driver-assist suite. The evaluators argued that the term “Autopilot” was misleading and irresponsible on Tesla’s part, and this was heavily taken against the Model 3’s rankings in the Assisted Driving Grading system.
“When it comes to the first test criterion – consumer information – the Tesla Model 3 in particular fails. The assistance systems are referred to as “Autopilot” in the operating instructions for the Model 3 as well as in the sales brochures and in marketing. However, the term suggests capabilities that the system does not have in sufficient measure. It tempts the driver to rely on the capabilities of the system – which is currently not allowed by the legislature anyway. Due to its good quick-start operating aid, the Tesla Model 3 still receives 10 points,” the evaluators noted.
Ultimately, these complaints about Autopilot’s branding ended up pulling down the Model 3’s scores to the point where the all-electric sedan was ranked below the Ford Kuga. Thatcham Research Director of Insurance Research Matthew Avery explained this in a video released about the evaluation. “The Tesla Model 3 was the best for safety backup and vehicle assistance but lost ground for misleading consumers about the capability of its Autopilot system and actively discouraging drivers from engaging when behind the wheel,” Avery said.
As noted by Avery, it is pertinent for vehicles to exhibit a balance to score very well in the Assisted Driving Grading system. This was not achieved by the Model 3 despite its industry-leading backup safety systems and actual vehicle assistance tech. ADAC explained it best when outlining why the Tesla Model 3 lost to four other vehicles despite being equipped with what is noticeably the most advanced driver-assist system.
“When analyzing the test results, it is noticeable that the Tesla Model 3 has the most advanced assistance systems. With 95 points for emergency assistance (Safety Backup) and 91 points for technical assistance, it doesn’t beat the Mercedes GLE by far, but at least 11 points… Because Euro NCAP removes the many points in the area of driver support from the Tesla, because on the one hand it does not sufficiently comply with the driver’s request for a steering correction. On the other hand, because Tesla is irresponsible about the term autopilot – an even more serious reason. With only 36 points from the test area driver integration, the Tesla falls back to sixth place in the final bill,” the ADAC noted.
Thatcham Research’s overall findings could be viewed in the video below.
SpaceX has disclosed the full financial details of its groundbreaking agreement with Anthropic, confirming that the AI company will pay $1.25 billion per month for dedicated high-performance computing resources.
The revelation came through SpaceX’s latest securities filing in preparation for its initial public offering, shedding light on one of the largest compute deals in the artificial intelligence sector to date. The prospectus was released last night, as SpaceX is heading toward its IPO.
This arrangement underscores the fierce demand for specialized infrastructure as frontier AI models require unprecedented levels of processing power to train and operate effectively. Industry analysts see the disclosure as a significant milestone, highlighting how top AI labs are locking in massive capacity to stay ahead in a rapidly accelerating field.
For SpaceX, it feels like a massive move that pushes its perception as a company from space exploration to artificial intelligence.
SpaceX is following in Tesla’s footsteps in a way nobody expected
The comprehensive deal grants Anthropic exclusive access to SpaceX’s Colossus clusters, encompassing Colossus I and the substantially expanded Colossus II, which together deliver hundreds of megawatts of power along with more than 200,000 NVIDIA GPUs.
Payments extend through May 2029, totaling nearly $45 billion overall; capacity is scheduled to ramp up during May and June 2026 at an initial discounted rate to facilitate seamless integration. Both companies retain the option to terminate the agreement with ninety days’ notice, so there is definitely some flexibility for both.
This pact not only enhances Anthropic’s ability to scale usage limits for Claude users but also injects substantial recurring revenue into SpaceX, bolstering its expansion into advanced data center operations and future orbital computing initiatives.
Observers describe the collaboration between the two companies as strategically advantageous because it gives Anthropic cutting-edge AI development the opportunity to collaborate with SpaceX’s expertise in rapid, large-scale infrastructure deployment.
This disclosure arrives at a pivotal moment when computing resources have become the primary bottleneck for AI progress.
As leading organizations compete to build more powerful systems, securing reliable, high-density facilities has emerged as a key differentiator.
SpaceX’s sites, such as those in Memphis, offer superior power availability and advanced cooling solutions that set them apart from conventional providers. For Anthropic, the added capacity is expected to deliver tangible improvements, including extended context windows, quicker inference times, and innovative features that appeal to both enterprise clients and individual users.
Looking ahead, the partnership paves the way for ambitious joint projects, including potential space-based AI compute platforms designed to overcome terrestrial limitations on energy and thermal management. Such efforts could redefine sustainable computing at massive scales.
Financially, the deal solidifies SpaceX’s diverse revenue profile ahead of its public market debut, extending beyond traditional aerospace activities. The massive check SpaceX will cash each month opens up the idea that additional
While some experts question the sustainability of these enormous expenditures given ongoing efficiency gains in AI architectures, the commitment reflects a strong belief in sustained demand growth.
The agreement also exemplifies productive synergies across sectors, with aerospace engineering insights optimizing AI hardware performance. As global attention on technology concentration increases, arrangements of this nature may help shape equitable access to critical resources.
Elon Musk
SpaceX just filed for the IPO everyone was waiting for
SpaceX filed its public S-1, revealing $18.7 billion in revenue and billions in losses.
SpaceX publicly filed its S-1 registration statement with the Securities and Exchange Commission on May 20, 2026, making its financial details available to the public for the first time ahead of what could be the largest IPO in history.
An S-1 is the formal document a company must submit to the SEC before going public. It includes audited financials, risk factors, business descriptions, and how the company plans to use the money it raises. Companies are required to file one before selling shares to the public, and it must be published at least 15 days before the investor roadshow begins. SpaceX had already submitted a confidential draft to the SEC in April, which allowed regulators to review the filing privately before it went public.
The S-1 reveals that SpaceX generated $18.7 billion in consolidated revenue in 2025, driven largely by its Starlink satellite internet division, which posted $11.4 billion in revenue, growing nearly 50% year over year. Despite that growth, the company lost about $4.9 billion in 2025 and has burned through more than $37 billion since its founding.
SpaceX just forced Verizon, AT&T and T-Mobile to team up for the first time in history
A significant portion of those losses trace back to xAI, Elon Musk’s artificial intelligence company, which was recently merged into SpaceX. SpaceX directed roughly 60% of its capital spending in 2025 to its AI division, totaling around $20 billion, yet that division lost billions and grew revenue by only about 22%.
SpaceX plans to list its Class A common stock on Nasdaq under the ticker SPCX, with Goldman Sachs, Morgan Stanley, and Bank of America leading the offering. The dual-class share structure means going public will not meaningfully reduce Musk’s control, as Class B shares he holds carry 10 votes per share compared to one vote for public Class A shares.
The company is targeting a raise of around $75 billion at a valuation of roughly $1.75 trillion, which would make it the largest IPO ever. The investor roadshow is reportedly planned for June 5.
Tesla has scaled back driver monitoring to be less naggy with the latest version of the Full Self-Driving (Supervised) suite, which is version 14.3.3.
The latest version is already earning praise from owners, who are reporting that the suite is far less invasive when it comes to keeping drivers from taking their eyes off the road. The first to mention it was notable Tesla community member on X known as Zack, or BLKMDL3.
14.3.3 nags less too https://t.co/IuiWzuYO6O
— Elon Musk (@elonmusk) May 18, 2026
Musk confirmed that v14.3.3 was made to nag drivers significantly less, something that Tesla has worked toward in the past and has said with previous versions that it is less likely to push drivers to look ahead, at least after looking away for a few seconds.
This refinement aligns with Tesla’s ongoing push toward unsupervised FSD. The update also brings faster Actual Smart Summon (now up to 8 mph), reliable “Hey Grok” voice commands, richer visualizations, smoother Mad Max acceleration, and an intervention streak counter that rewards consistent use. Reviewers describe the drive as more human-like and confident, with fewer twitches or unnecessary maneuvers.
Musk has repeatedly signaled this direction. In late 2025, he stated that FSD would allow phone use “depending on context of surrounding traffic,” noting safety data would justify relaxing rules so drivers could text in low-risk scenarios like stop-and-go traffic.
We tested this, and even still, the cell phone monitoring really seems to be less active in terms of alerting drivers:
Tesla Full Self-Driving v14.2.1 texting and driving: we tested it
Earlier, ahead of v14, Musk promised the system would “nag the driver much less” once safety metrics improved.
In 2023, he confirmed the steering wheel torque nag would be “gradually reduced, proportionate to improved safety,” shifting reliance to the cabin camera. Subsequent updates like v13.2.9 and v12.4 further loosened monitoring, cracking down on workarounds while easing legitimate distractions.
These steps reflect Tesla’s data-driven approach: FSD’s safety record—reportedly averaging millions of miles per crash—now outpaces human drivers in many scenarios, giving the company confidence to dial back interventions. Reduced nags improve usability and trust, encouraging more drivers to rely on the system rather than disengaging out of frustration.
However, there are certainly still some concerns. In many states, it is illegal to handle a cell phone in any way, requiring the use of hands-free devices. In Pennsylvania, it is illegal to use your cell phone at stop lights, which is definitely a step further than using it while the car is actively in motion.
v14.3.3 represents tangible progress. Making FSD less adversarial and more seamless is definitely a step forward, but drivers need to be aware of the dangers of distracted driving. FSD is extremely capable, but it is in no way fully autonomous, nor does its performance warrant owners to take their attention off the road.
SpaceX reveals what Anthropic will pay for massive compute deal
SpaceX just filed for the IPO everyone was waiting for
