According to monitoring by Beating, SenseTime Chairman and CEO Xu Li said in a keynote speech at the 2026 World Artificial Intelligence Conference that AI accessibility and AI safety need to be addressed together.
Xu said several AI applications already validated by the market — software development, content generation and embodied intelligence — are moving along a similar track. In his description, they are shifting from single-module tools built for professionals to long-horizon task closed loops for the general public.
He said the clearest change in that process is the migration of users. The original target customer is no longer the key focus, while the real customer base is emerging naturally from the demand side.
AI accessibility should center on augmentation
Xu said real AI accessibility is not about replacing humans. He framed it instead as expanding the boundaries of capabilities that individuals did not previously possess.
At the product level, he said AI should be designed for augmentation rather than replacement. It also should not deepen dependence, but support personal growth.
Safety governance spans products, models and physical boundaries
On governance, Xu said AI safety needs to run through three layers: product philosophy, model pathways and physical boundaries.
At the model layer, he said data security should be ensured during both pre-training and post-training, while multi-model cross-verification should be used in deployment. For embodied intelligence, he said rollout should proceed gradually through sandboxes and scenario constraints.
Pricing unit may move from token to task
Xu also said that as AI capability advances toward what he called the era of the super individual, the real unit of pricing in the future will no longer be the token, but the price of completing a task.
He said the Task economy will push AI away from pure compute consumption and toward value delivery. As costs keep falling, the threshold for broader access will also open further.
Quoting World Economic Forum data, Xu said AI could reduce 90 million jobs by 2030, while creating 170 million new professions. He said super individuals will sit at the center of that new form of work.