Waymo releases a self-driving open data set with 1,000 segments of 20-second footage in a range of driving conditions, free for non-commercial research
Darrell Etherington / TechCrunch :
Context & Ripple Effects
Waymo's release follows UC Berkeley's move a year earlier to open-source what it called the world's largest autonomous driving dataset, establishing a pattern of research-grade data sharing in self-driving. It also sits alongside earlier coverage of how Waymo and Tesla collect data from billions of driven miles — this dataset is a curated slice of that private trove, gated to non-commercial use.
The arc matters because Waymo's openness is selective and directional: after this research set came published Arizona road-testing data covering 6.1M+ miles driven in 2019, while by 2022 Waymo was suing the California DMV to keep crash and safety details designated as trade secrets. The company shares perception data freely but guards operational safety records tightly.
First-order effects
- Non-commercial researchers gain labeled real-world driving footage across varied conditions without needing their own fleet, lowering the entry cost for perception and prediction work at universities and labs.
- Waymo gets a recruiting and standards lever: academic teams benchmarking on Waymo's format align their tooling with the company's sensor suite rather than competitors'.
Second-order effects
- Rivals and suppliers face pressure to match the release cadence or explain why their data stays closed, since Berkeley's and Waymo's sets become the de facto comparison baselines for published research.
Third-order effects
- If the pattern holds, autonomy splits into two disclosure regimes — perception datasets shared openly for research while incident and safety telemetry is walled off as proprietary, forcing regulators to negotiate access case by case, as the California DMV fight shows.
The trend: Self-driving leaders are normalizing curated open datasets for academia while progressively hardening legal protection around operational safety data.