7 hrs ago
AI Development Keeps Racing Despite Calls for Safety
People who build AI have warned that developing it too quickly could create safety problems.
But companies in the United States and China are still releasing new AI models.
The article says Chinese companies launched 16 models in September.
US labs launched or updated at least 10 to 12 models and related versions that month.
Some AI models have open weights, which lets developers run and change them themselves.
Researchers worry that these models could be misused or that their safety protections might fail.
The article says these concerns apply to companies around the world.
It argues that competition makes companies focus on speed and cost, while binding global safety rules have not yet been established.
AI researchers and company leaders have warned that rapid development poses safety risks, but the article reports no sign of a slowdown.
Donald Trump has argued that US frontier labs cannot afford to slow down, saying, “whoever wins AI, wins,” and citing competition with China.
A Nikkei tabulation says 10 leading Chinese AI companies launched 16 models in September.
Major US labs launched or updated at least 10–12 models, variants, and specialized reasoning sub-tiers that month.
The article says open-weight risks and the tension between rapid releases and safety checks affect the global AI industry, not only Chinese developers.
- Who
- AI companies and researchers in the United States and China, with safety concerns also raised by Anthropic and global security researchers.
- What
- AI model development and releases continue despite warnings about safety and calls to slow down.
- Where
- The United States and China, within the wider global AI industry.
- When
- The article cites model launches and updates in September; it does not specify a year.
- Why
- The article says competition, speed, and cost-efficiency drive continued development, while global binding governance standards have not been established.
Safety and verification
Competition and continued development
Whether development should slow
Safety and verification
AI pioneers, safety researchers, and some AI CEOs have called for slowing rapid development to address safety and risks to humanity.
Competition and continued development
Donald Trump has argued that US frontier labs cannot afford to slow down because “whoever wins AI, wins,” particularly amid competition with China.
Open-weight models
Safety and verification
Anthropic and security researchers warn about risks including automated exploit generation, guardrail failures, and possible misuse of open-weight releases.
Competition and continued development
The article notes that high-capability open-weight models are also released by US companies and French lab Mistral, and says the safety dilemmas are not exclusive to Chinese developers.
Key facts
- Chinese company tally
- Nikkei's tabulation says 10 leading Chinese AI companies launched 16 models in September.
- US model activity
- Major US labs launched or updated at least 10–12 models, variants, and specialized reasoning sub-tiers in September.
- Chinese companies named
- DeepSeek, Alibaba/Qwen, Moonshot AI, Zhipu/GLM, Tencent, and Xiaomi.
- US labs named
- OpenAI, Anthropic, Google, SpaceXAI, and Meta.
- Open-weight models
- The article defines these as models whose underlying weights developers can access, customize, and run independently.
- Distillation
- The article describes it as using a larger teacher model to train a smaller, faster model.
- Safety concerns
- Researchers cite automated exploit generation, guardrail failures, and cross-border model distillation.
Quotes
Donald Trump
US president cited in the article discussing competition in AI development.
“whoever wins AI, wins”
NDTV
Sources
Related news
Anthropic CEO Urges AI Slowdown Amid Safety and Monopoly Concerns
Former Researcher Warns Chinese Spies May Have Infiltrated AI Labs
Trump Rejects AI Slowdown, Says Whoever Wins AI Wins
AI Safety Debate Grows Over Pacing Frontier Development
China Joins Trump in Debate Over AI Development Pace
Trump Rejects AI Slowdown as US Races Ahead of China
Anthropic Consults Religious Leaders on Claude’s Values and Consciousness
Global AI Governance Splits as Nations Debate Voluntary Safeguards
Google Unveils Gemini 4 Argon for Long-Horizon AI Workflows
October 2026 AI Updates Center on Agents, Safety and Models
