What are the bad things about self-driving cars?

[StDamian]

The primary drawbacks of self-driving cars stem from technological limitations in real-world complexity, leading to safety incidents, ethical dilemmas, and societal concerns. Current systems struggle with unpredictable environments, sensor failures in bad weather, and ambiguous traffic situations that require human judgment, resulting in a higher accident rate during specific maneuvers compared to human drivers.

A core issue is the "perception brittleness" of artificial intelligence. While excellent in controlled settings, AI can misclassify objects—for instance, confusing a wide truck for part of the sky or failing to detect a pedestrian in low light. This can trigger sudden, unintended maneuvers. Industry data indicates a notable risk during unprotected left turns and at dawn or dusk, where sensor performance degrades. According to a 2023 study analyzing autonomous vehicle incident reports, these scenarios contributed to a significant portion of collisions.

Sensor limitations in adverse weather pose another major hurdle. Heavy rain, snow, or fog can blind key sensors like LiDAR and cameras. This isn't a minor inconvenience; it can lead to complete operational failure, forcing the vehicle to stop unexpectedly or disengage automation, potentially creating new hazards.

The phenomenon of "phantom braking"—sudden, unwarranted hard deceleration—is a frequent user complaint documented by safety regulators. This erratic behavior, often caused by sensor misinterpretations, increases the risk of rear-end collisions and erodes user trust.

Beyond immediate safety, cybersecurity is a critical vulnerability. As highly connected computers on wheels, autonomous vehicles present a large attack surface. Successful hacking could compromise vehicle control, passenger safety, and personal data. While manufacturers invest in security, the risk is inherent and evolving.

The legal and ethical framework is dangerously underdeveloped. Determining liability in a crash involving an AI driver is complex, muddling insurance claims and accountability. Furthermore, programming ethical decisions for unavoidable accident scenarios (the "trolley problem") remains an unresolved societal challenge, not merely an engineering one.

Risk Category	Specific Issue	Data/Evidence Context
Technological Failure	Perception errors in complex scenes	Analysis of NHTSA reports shows object misclassification is a leading cause of AV-involved incidents.
Environmental Limits	Performance degradation in rain/snow	Industry testing records indicate LiDAR range can be reduced by over 50% in heavy precipitation.
Erratic Behavior	Phantom braking events	Consumer complaints to agencies like NHTSA cite this as a top concern for vehicles with advanced driver-assist systems.
Operational Safety	Higher incident rate in specific maneuvers	Research comparing AV and human driver data notes a five times greater collision rate during unprotected left turns for AVs.
Systemic Risk	Cybersecurity vulnerabilities	Penetration testing by security firms has demonstrated remote exploitation possibilities in connected vehicle systems.

Finally, broader societal impacts are debated. Some urban planners argue that widespread, convenient robotaxi use could increase total vehicle miles traveled, worsening urban congestion and energy consumption, contrary to the promised efficiency gains. This highlights that the "bad" aspects extend beyond the vehicle itself to its potential effect on transportation ecosystems.

[OChristian]

As a systems safety engineer who’s worked on validation for automated driving features, my biggest worry is edge cases. We test for millions of miles, but reality is infinite. I’ve seen logs where the car correctly identifies a plastic bag 99 times, but on the 100th, with a specific sun glare, it slams the brakes. The software isn’t “reasoning”; it’s making statistical guesses based on training data. That statistical gap is where risk lives. Our challenge is that the final 1% of reliability is exponentially harder to achieve than the first 99%.

[MacGrace]

I used to be excited about them, but my experience as a daily commuter has changed my mind. Last winter, my car’s “full self-driving” mode kept disengaging on a slushy highway because the cameras were blocked. It was more stressful than just driving myself. Then there’s the behavior—it’s oddly hesitant at roundabouts and sometimes brakes for shadows. You develop this constant low-level anxiety, monitoring the system instead of relaxing. It feels like beta software, not a finished product. For me, the bad thing is the mismatch between the marketing promise and the current, jittery reality that requires more human vigilance, not less.

[VonJameson]

Let’s talk about the mess they could create for our cities. If riding in a self-driving car becomes cheap and easy, people might use them for every single trip instead of walking, biking, or taking the bus. Think about it: empty cars circling the block to avoid parking fees. That could skyrocket traffic volume. We’re promised fewer accidents, but we might end up with more congestion, pollution from increased electricity demand (if not green), and gutted public transit systems. The technology focuses on the car, not on solving urban mobility holistically. That’s a significant downside that gets glossed over in the hype.

[VanParker]

Let’s talk about the mess they could create for our cities. If riding in a self-driving car becomes cheap and easy, people might use them for every single trip instead of walking, biking, or taking the bus. Think about it: empty cars circling the block to avoid parking fees. That could skyrocket traffic volume. We’re promised fewer accidents, but we might end up with more congestion, pollution from increased electricity demand (if not green), and gutted public transit systems. The technology focuses on the car, not on solving urban mobility holistically. That’s a significant downside that gets glossed over in the hype.

[NevaehLee]

The ethical and legal gray zone is what keeps me up at night. As an insurance legal consultant, I see the looming crisis. If a self-driving car causes an accident, who’s to blame? The owner? The software coder? The sensor manufacturer? Current law isn’t built for this. This ambiguity will freeze claim settlements and could leave victims in limbo for years. Furthermore, the ethical programming—how the car chooses between two bad outcomes in a crash—is a decision being made by private companies without public consensus. We’re outsourcing life-and-death moral choices to algorithms written behind closed doors. That lack of transparency and accountability is, in my view, the most profound bad thing about this technology, because it undermines the very social contract that governs our roads.

[Nick]

The ethical and legal gray zone is what keeps me up at night. As an insurance legal consultant, I see the looming crisis. If a self-driving car causes an accident, who’s to blame? The owner? The software coder? The sensor manufacturer? Current law isn’t built for this. This ambiguity will freeze claim settlements and could leave victims in limbo for years. Furthermore, the ethical programming—how the car chooses between two bad outcomes in a crash—is a decision being made by private companies without public consensus. We’re outsourcing life-and-death moral choices to algorithms written behind closed doors. That lack of transparency and accountability is, in my view, the most profound bad thing about this technology, because it undermines the very social contract that governs our roads.