Tesla Unveils Automatic Collision Evasion: FSD Steps In When Drivers Don't
Tesla has quietly introduced a new active safety feature called Automatic Collision Evasion that can engage its Full Self-Driving (Supervised) software to take evasive action even when the driver is operating the vehicle manually. The feature, first reported by Not a Tesla App on September 3, 2026, marks a significant expansion of Tesla's autonomy stack from a driver-assist mode to an always-on safety net.
According to the report, the system activates in two specific scenarios. The first is when an imminent frontal collision is detected and braking alone cannot prevent it. In such cases, the vehicle will automatically enable FSD to steer around the obstacle—changing lanes or moving onto the shoulder—while modulating speed. This intervention is limited to controlled-access highways at speeds under 85 mph (137 km/h), and it only activates if no pedestrians, cyclists, or vulnerable road users are in the intended path, and the road surface is dry.
The second trigger addresses driver inattention. If the vehicle determines that the driver is no longer in control—such as when reaching into the back seat or looking away—and the car is traveling between 22.5 mph and 85 mph on any roadway type, it will automatically assume control and guide the vehicle to safety.
Currently, Automatic Collision Evasion is available only in North America. Tesla has not announced a timeline for other regions, including the Netherlands, which became the first EU country to approve FSD earlier in 2026.
How It Works: Overriding the Driver for Safety
The system is designed to work alongside existing safety features like Automatic Emergency Braking (AEB), Obstacle-Aware Acceleration, and blind-spot monitoring. But unlike AEB, which typically brakes in a straight line, Automatic Collision Evasion can make proactive steering maneuvers. Tesla's neural networks have demonstrated the ability to predict and avoid crashes at highway speeds when FSD is engaged. This new feature extends that capability to manual driving, effectively making the autonomy stack a passive safety watchdog.
The introduction raises important questions about driver override protocols. While Tesla has not released full technical specifications, the feature appears to take control temporarily, then hand it back once the danger has passed. The company has not specified how the system handles situations where the driver is actively steering but the system disagrees, or what happens if the driver resists the intervention.
The Context: Tesla's Safety Claims vs. Crash Data
Tesla's move comes amid intense scrutiny over the safety of its driver-assist systems. In recent weeks, an investigation by Electrek has uncovered a pattern of fatal crashes involving Autopilot and FSD where Tesla's own data confirms the systems were active, yet public reports often omit this critical detail.
Fatal Crash in Clute, Texas
On August 31, 2026, Electrek reported that Tesla confirmed to federal regulators that its driver-assist system was "Verified Engaged" in a crash that killed 23-year-old Steven Alvarez in Clute, Texas. The crash, which occurred on May 20, 2026, involved a 2025 Model 3 that left Brazoswood Drive, crashed through a fence at Clute Municipal Park, and struck a decommissioned pool complex. Tesla's report to the National Highway Traffic Safety Administration (NHTSA) recorded a pre-crash speed of 104 mph on a road categorized as a "Street," and listed the outcome as a fatality.
However, Tesla redacted nearly all explanatory details—the crash narrative, software version, and whether the road was within the system's approved operating area—as "confidential business information." This redaction is standard in 99.9% of Tesla's NHTSA filings, according to Electrek's investigation.
Fatal Left-Turn in Batavia, Illinois
A second report revealed that FSD was engaged in a collision on March 25, 2026, in Batavia, Illinois. A Tesla Model Y attempting a left turn at an intersection struck a Dodge pickup truck traveling straight, killing front-seat passenger Maggie Espinosa, a 37-year-old mother of five. Tesla's NHTSA filing lists the pre-crash movement as "Making Left Turn" at 24 mph, with the system "Verified Engaged." Again, the software version and other key details were redacted.
This case points directly to FSD rather than basic Autopilot, as the maneuver was an unprotected left turn on city streets—a feature unique to FSD.
Stalled on the Freeway in Mesa, Arizona
Even more concerning is a pattern of Teslas stopping in live freeway lanes with driver-assist engaged. A crash on October 31, 2025, in Mesa, Arizona, killed a driver whose 2020 Model 3 was rear-ended while stopped in a freeway lane. Tesla's report confirms the system was "Verified Engaged" and the car was "Stopped" at 0 mph. A similar incident occurred weeks before in Florida, although details remain scant.
These cases have led Electrek to characterize Tesla's redaction practices as an attempt to "hide" the role of its systems, though Tesla argues the information is protected under confidential business information exemptions.
The Stakes: Trust, Regulation, and the Future of Autonomous Safety
The rollout of Automatic Collision Evasion comes at a critical juncture for Tesla and the broader autonomous vehicle industry. On one hand, the feature represents a potential leap forward in crash avoidance, using advanced neural networks to predict and prevent accidents that human drivers might not react to in time. On the other hand, the recent crash data underscores the limitations and risks of relying on the same technology to intervene in emergencies.
Regulatory and Public Trust Challenges
NHTSA's Standing General Order requires automakers to report crashes involving Level 2 driver-assist systems, but Tesla's heavy redaction limits public understanding. Unlike competitors such as Ford and General Motors, which often provide detailed crash narratives, Tesla's reports are largely black boxes. This lack of transparency makes it difficult for regulators and consumers to assess the true safety record of FSD.
The new Automatic Collision Evasion feature may face additional regulatory scrutiny. Since it operates without driver input, even during manual driving, it blurs the line between Level 2 (driver-supervised) and Level 3 (conditionally automated) systems. If the system performs evasive maneuvers without driver consent, it could be argued that Tesla is testing higher levels of automation without the necessary regulatory approvals.
Technological and Ethical Considerations
The feature's activation criteria reveal Tesla's caution: it only operates under ideal conditions—dry roads, no vulnerable road users, and speeds below 85 mph. This suggests the company is aware of the system's limitations. Yet, the recent crashes show that FSD can fail in unexpected ways, such as misjudging a left-turn or stopping inexplicably in a freeway lane.
The development also raises ethical questions about intervention. Should a car override a driver's actions to avoid a collision? How does the system weigh the risk of a maneuver against the risk of staying course? Tesla has not published its decision-making algorithms, but the feature's existence signals a growing confidence in AI-driven safety systems.
What's Next for Tesla and the Industry
Tesla is not alone in exploring automatic collision avoidance. Most major automakers offer some form of autonomous emergency braking with steering assist, but few have the capacity to leverage a full self-driving stack for such interventions. If Automatic Collision Evasion proves effective, it could set a new standard for active safety and pressure competitors to adopt similar technologies.
However, public trust remains fragile. As Electrek's investigation continues to uncover discrepancies between Tesla's data and official crash reports, calls for greater transparency are likely to intensify. Regulators may demand that Tesla reveal more information about when and how its systems are engaged, especially if new features like Automatic Collision Evasion become commonplace.
For now, Tesla owners in North America with compatible hardware will receive the feature as part of an over-the-air update. The company has not yet detailed how the system will behave in all situations, or whether it can be disabled. But the message is clear: Tesla sees its autonomous technology as not just a convenience, but a life-saving safety net.
As the industry watches, one thing is certain—the line between driver and machine is becoming more blurred than ever. And with each passing month, the need for clear, unbiased data on how these systems perform in the real world becomes more urgent.
Comments