China's AI gains are testing America's strategy
Image: Just Security
Chinese startup Zhipu AI has released a new AI model that can match Anthropic's flagship Mythos system at finding software security bugs in certain benchmark tests, according to cybersecurity company Semgrep.
While the model, called GLM-5.2, still trails America's top AI systems in many broader tasks, it's another sign the technology gap between the two countries is shrinking.
- Zhipu's GLM-5.2 is open-weight, unlike Anthropic and OpenAI's newest models, meaning companies can download, modify, and run it on their own hardware.
- That appeals to businesses looking for lower-cost AI and more control over sensitive data, but it also gives hackers access to the same bug-finding capabilities without the safeguards built into most closed AI systems.
Security teams can use them to find hidden software flaws before they're exploited, review huge amounts of code in a fraction of the time, and help developers fix security problems sooner.
But at the same time, bad actors can use the latest AI models to look for those same weak spots before they're patched, search thousands of programs at once for potential entry points, and automate work that once took teams of people.
Striking an AI balance
The potential downsides of new models have drawn pushback from the White House in recent weeks.
- The Trump admin earlier this month blocked broad access to Anthropic's Mythic 5 model, before restoring it for certain trusted organizations under a deal reached last Friday.
- That same day, OpenAI limited access to its newest GPT-5.6 models to a small group of customers approved by the Trump admin, citing discussions with US officials reviewing their cybersecurity implications.
Some researchers worry the current US strategy could backfire. Chinese companies continue releasing cheaper, increasingly capable open-weight models while companies including Microsoft explore offering them on their platforms. If those models become easier for businesses to access, America's AI advantage may depend as much on adoption as innovation.
Zoom out: AI excitement is still running high, but some investors on Wall Street are becoming more selective about where they're placing their bets.
OpenAI is reportedly leaning toward delaying its IPO until next year as advisers question whether investors are ready to support CEO Sam Altman's targeted $1 trillion valuation, while the company continues spending billions on data centers, computing power, and recruiting top AI talent.
📊 Flash poll: In your opinion: Should the US govt’t have fewer rules around AI to promote innovation and boost competitiveness globally, even if it carries increased risks? Or should the US have stricter rules to make AI safe and/or fair, even if it slows innovation and competitiveness globally?
See a 360° view of what pundits are saying →
Sprinkles from the Left
- Some commentators argue that the Trump administration risks weakening America's lead in AI by placing unpredictable restrictions on domestic AI companies, and that a more balanced approach is needed to protect national security without slowing innovation or driving talent elsewhere.
- Others contend that growing competition from open-source AI models could make it harder for companies like Anthropic and OpenAI to justify their premium prices. They argue that as the technology becomes more widely available, long-term success will depend on offering clear value beyond the models themselves.
- “Can America Avoid a Jack Ma Moment?” –Dan Wang and Julian Gewirtz, New York Times
- “Competition intensifies for Anthropic and OpenAI ahead of IPOs” –Richard Waters, Financial Times
- “America’s lead in AI is now at the mercy of local zoning boards” –Warren Wimmer, Chattanooga Times Free Press
Sprinkles from the Right
- Some commentators argue that maintaining US leadership in AI will require stronger safeguards against Chinese technological competition, and that finding a way to protect advanced AI while balancing innovation with national security is essential to preserving America’s global advantage.
- Others contend that the Trump administration's restrictions on Anthropic's AI models risk undermining American innovation, warning that unpredictable government intervention could weaken the nation's technological edge and make it harder for US companies to stay ahead of global competitors.
- “How America can remain the world’s AI superpower” –Hal Brands, Washington Post
- “Why Does Trump Hate Anthropic?” –Wall Street Journal Editorial Board
- “Trump’s quantum plan is a needed response to the foreign AI threat” –Sam Korkus, Washington Examiner
Share this!
Recent Discussion stories
Discussion
| June 26, 2026Trump delays bipartisan housing bill over SAVE Act
After clearing both the Senate and House this week, the bipartisan Road to Housing Act appeared to be on track to become law.
Discussion
| June 24, 2026US gov’t joins nationwide push for nuclear energy
The Trump administration announced Tuesday that the Department of Energy will provide billions of dollars in loans to help accelerate construction of nuclear reactors across the US.
Discussion
| June 22, 2026Iran pauses peace talks with the US
The Trump administration's newly signed peace agreement with Iran is already facing major headwinds.
You've made it this far...
Let's make our relationship official, no 💍 or elaborate proposal required. Learn and stay entertained, for free.👇
All of our news is 100% free and you can unsubscribe anytime; the quiz takes ~10 seconds to complete