跳到内容

State of AI 2026

Risks & Pain Points

AI Risks

The AI industry has never been shy about discussing the technology's potential for disaster, sadly stopping just short of actually, you know, slowing down for a minute.

Said risks include job displacement as a top concern, but also AI's potential for military use as seen in recent conflicts, and of course the worrying environmental impact of AI in a world already threatened by climate change.

Many commenters complained about being limited to selecting only three items. As one person put it: “It's kind of shocking how big this list of negative social impacts are when you put them all in one place.”

Which of these general AI risks and issues are you most concerned about?
多选
0%
20%
40%
60%
80%
100%
01
Job displacement
3,003
02
Military use of AI
2,804
03
Environmental impact
2,490
04
AI slop takeover
2,107
05
Negative cognitive impacts
1,850
06
Security issues
1,821
07
AI overuse
1,106
08
Rising AI costs
1,031
09
Copyright and ethics issues
730
10
Open-source disruption
597
0%
20%
40%
60%
80%
100%
受访者百分比
6,257
86%
10

痛点

When it comes to issues with actually using AI, Hallucinations lead the pack.

What makes this issue uniquely frustrating is not just having to deal with incorrect results, but incorrect results confidently presented as correct, with the model having no awareness of its own potential for error.

在使用 AI 工具时，您最大的痛点是什么？
多选
0%
20%
40%
60%
80%
100%
01
幻觉和不准确
+8
3,899
02
代码质量
-1
3,249
03
缺乏上下文
-1
2,321
04
财务成本
+3
1,923
05
隐私担忧
-1
1,903
06
提示词理解
-3
1,496
07
版权担忧
-2
784
08
集成困难
+1
784
09
响应时间慢
10
道德伦理担忧
11
缺乏对提示词的理解
12
生成的代码质量差
13
其他答案
173
0%
20%
40%
60%
80%
100%
受访者百分比
6,084
84%
10

缺失的功能

Confirming the previous question's results, Truthfulness is the leading missing feature. In an age where we're growing increasingly skeptical of our fellow humans' ability to distinguish fact from fiction, it'd be nice if our AI at least wouldn't gaslight us as well.
您觉得 AI 工具目前缺少哪些功能？
多选
0%
20%
40%
60%
80%
100%
01
真实性
2,801
02
Long-term memory
2,592
03
最新的知识
2,032
04
Accountability
1,977
05
Awareness of runtime behavior
1,345
06
High-level planning
1,320
07
更大的上下文窗口
1,151
08
更好的 UI/UX
1,018
09
支持更大的代码库
1,007
10
更好的工具集成
628
0%
20%
40%
60%
80%
100%
受访者百分比
5,859
81%
10

满意度

TODO
您对 Web 开发 AI 工具的现状满意吗？
1
2
3
4
5
2025
2026
1
2
3
4
5
平均 {axis}
6,212
86%
3.4
4
10

资源推荐

MCP: Learn the Universal Language for AI Integration
Brian Holt
Neon

MCP: Learn the Universal Language for AI Integration

Dive into the Model Context Protocol and code an MCP server from scratch! Leverage external APIs, interact with custom applications, and learn security best-practices to ensure safe AI agent deployment.
Hard Parts of Neural Networks
Will Sentance
Codesmith

Hard Parts of Neural Networks

Through hands-on examples such as fraud detection and image recognition, you'll build robust mental models for foundational AI concepts like data preprocessing, probability, training, and prediction.
感谢合作伙伴对我们的支持! 了解更多。