Tesla Establishes Ambitious 10 Billion Mile Target for Unsupervised Full Self-Driving; Ashok Clarifies the Justification Behind This Initiative
In a recent update on X, Elon Musk outlined the criteria Tesla considers essential for attaining superhuman safety standards for its Unsupervised Full Self-Driving (FSD) system. Furthermore, Tesla's Director of AI, As...
In a recent update on X, Elon Musk shared Tesla's updated requirements for achieving superhuman safety levels for Unsupervised Full Self-Driving (FSD). Tesla's Director of AI, Ashok Elluswamy, also confirmed that some reasoning capabilities—essential for advanced AI decision-making—have already been integrated into FSD, with more enhancements on the way.
**10 Billion Miles: A New Benchmark**
In Elon’s Master Plan Part Deux, he mentioned a critical figure—6 billion miles—representing the data necessary for securing global regulatory approval for true autonomy through Unsupervised FSD. This milestone was aimed at creating a system that is significantly safer than human drivers, systematically addressing edge cases and real-world complexities.
That target has now been updated to 10 billion miles of training data required for safe unsupervised self-driving. The challenge lies in the extensive complexity of real-world scenarios.
Elon and Tesla are focused on accumulating real-world miles. While they, like other autonomous competitors, utilize simulated data for training, actual driving data is crucial for navigating real-world edge cases effectively.
**Tesla's Current Progress**
As reported by Tesla’s FSD Safety Hub, the fleet has accumulated approximately 7.1 billion miles while using FSD, indicating that around 2.9 billion more miles are needed to meet the new target.
With a gradual increase in the FSD adoption rate and fleet growth each year, this goal appears attainable in the near future. Tesla recorded 3.6 billion miles on FSD in March 2025 and has now reached the current total of 7.1 billion. At the current pace, Tesla is likely to achieve the 10 billion-mile mark within the next six months.
Elon’s candid acknowledgment of the challenges associated with addressing edge cases underscores the rigorous journey toward achieving higher safety standards. While reaching 99% safety may seem straightforward, advancing to 99.9% and beyond becomes increasingly complex. This observation comes shortly after similar cautionary remarks made by Elon and Ashok to NVIDIA during the unveiling of Alpamayo at CES 2026.
The vast array of edge cases—from uncharted routes and sudden changes to unpredictable human or animal behaviors and adverse weather conditions—poses significant challenges.
**Advancements in Reasoning with V14**
To bridge the gap between 7 billion and 10 billion miles, Tesla is not merely increasing data volume; they are evolving how FSD processes decisions.
Ashok highlighted that FSD V14.2, currently deployed in customer vehicles, features initial implementations of reasoning capabilities. This includes functionalities such as adjusting navigation routes during construction and identifying parking options, with further advancements expected in Q1.
**Understanding Reasoning**
Historically, FSD has functioned primarily as a quick-response system—reacting to immediate stimuli like traffic signals or obstacles. The introduction of reasoning capabilities allows the vehicle to evaluate multiple potential future scenarios before making decisions, akin to human thought processes.
In brief, this transition to a more proactive approach facilitates more human-like driving. For instance, rather than abruptly stopping at a blocked road, the vehicle can now reason that it needs to reroute, assess its surroundings, and choose an alternative path.
In the context of parking, this means intelligently selecting suitable locations, such as finding a safe parking spot rather than simply stopping at the nearest destination.
**Key Insights**
Elon’s reference to the 10 billion-mile target reflects the complexities of reality, which have proven to be more challenging than initially anticipated at Tesla. However, Ashok’s update reassures that a solution—Vision-Language-Action models capable of thinking rather than merely reacting—is now within reach.
**10 Billion Miles: A New Benchmark**
In Elon’s Master Plan Part Deux, he mentioned a critical figure—6 billion miles—representing the data necessary for securing global regulatory approval for true autonomy through Unsupervised FSD. This milestone was aimed at creating a system that is significantly safer than human drivers, systematically addressing edge cases and real-world complexities.
That target has now been updated to 10 billion miles of training data required for safe unsupervised self-driving. The challenge lies in the extensive complexity of real-world scenarios.
Elon and Tesla are focused on accumulating real-world miles. While they, like other autonomous competitors, utilize simulated data for training, actual driving data is crucial for navigating real-world edge cases effectively.
**Tesla's Current Progress**
As reported by Tesla’s FSD Safety Hub, the fleet has accumulated approximately 7.1 billion miles while using FSD, indicating that around 2.9 billion more miles are needed to meet the new target.
With a gradual increase in the FSD adoption rate and fleet growth each year, this goal appears attainable in the near future. Tesla recorded 3.6 billion miles on FSD in March 2025 and has now reached the current total of 7.1 billion. At the current pace, Tesla is likely to achieve the 10 billion-mile mark within the next six months.
Elon’s candid acknowledgment of the challenges associated with addressing edge cases underscores the rigorous journey toward achieving higher safety standards. While reaching 99% safety may seem straightforward, advancing to 99.9% and beyond becomes increasingly complex. This observation comes shortly after similar cautionary remarks made by Elon and Ashok to NVIDIA during the unveiling of Alpamayo at CES 2026.
The vast array of edge cases—from uncharted routes and sudden changes to unpredictable human or animal behaviors and adverse weather conditions—poses significant challenges.
**Advancements in Reasoning with V14**
To bridge the gap between 7 billion and 10 billion miles, Tesla is not merely increasing data volume; they are evolving how FSD processes decisions.
Ashok highlighted that FSD V14.2, currently deployed in customer vehicles, features initial implementations of reasoning capabilities. This includes functionalities such as adjusting navigation routes during construction and identifying parking options, with further advancements expected in Q1.
**Understanding Reasoning**
Historically, FSD has functioned primarily as a quick-response system—reacting to immediate stimuli like traffic signals or obstacles. The introduction of reasoning capabilities allows the vehicle to evaluate multiple potential future scenarios before making decisions, akin to human thought processes.
In brief, this transition to a more proactive approach facilitates more human-like driving. For instance, rather than abruptly stopping at a blocked road, the vehicle can now reason that it needs to reroute, assess its surroundings, and choose an alternative path.
In the context of parking, this means intelligently selecting suitable locations, such as finding a safe parking spot rather than simply stopping at the nearest destination.
**Key Insights**
Elon’s reference to the 10 billion-mile target reflects the complexities of reality, which have proven to be more challenging than initially anticipated at Tesla. However, Ashok’s update reassures that a solution—Vision-Language-Action models capable of thinking rather than merely reacting—is now within reach.
