AI SAFETY TWITTER UPDATES

← Front page

Reportedly: China's MOFCOM to require approval for open-sourcing models above 2T params / 10^26 FLOPS

4 people · 11 posts · first post · latest

Posts circulating claim China's MOFCOM will require approval for Chinese labs to open-source models above 2T parameters or 10^26 FLOPS, widely read as restricting Chinese open-weight releases. Nathan Lambert notes an accompanying statement offering near-automatic licensing to support open-source AI, while Kyle Chan points to MOFCOM export-control thresholds for models' cyber capabilities, which he says aren't far from current models. Lambert warns most commentary parses partial AI translations of a short Chinese document; details remain unconfirmed.

All posts · 8, most important first

  1. @natolambert research ·
    Seems very important, as tons of people are tweeting about the end of bigger open models from China due to this from MOCOM, but there's a statement intending to make it easier for open source technology. Much hinges on implementation and how "near automatic" approval would work.
    MofCom said they will have a near auto licensing regime for open-source AI that allows more rapid release it seems? "To support the development of an open-source AI innovation ecosystem, automatic licensing and other facilitation measures will apply to technology exports involving open-source releases of AI technologies, Chinese companies building and operating their own computing centers overseas, overseas chip tape-outs and fabrication, joint R&D using overseas resources, and participation in international standard-setting—essentially allowing exports to proceed immediately upon filing." open on X
    8 likes · 2 reposts
  2. @_nathancalvin policy ·
    @kyleichan
    Can we please start reading the appendix of China MOFCOM’s export control list? If a model’s cyber capabilities exceed certain thresholds, it could fall under export controls. And in my view, those thresholds aren’t particularly far from current frontier capabilities. mofcom.gov.cn/zcfb/dwmygl/ar… open on X
    0 likes · 0 reposts
  3. @hamandcheese policy ·
    image
    If Chinese players want to open source models above 2T parameters or trained on over 10e26 FLOPS then they will need to get permission from MOFCOM open on X
    0 likes · 0 reposts
  4. @_nathancalvin policy ·
    Can @mattsheehan88 @kyleichan or other smart China AI watchers tell us how big a deal this MofCom stuff is? I am confused. What does/doesn't this mean for next gen Chinese open weight releases?
    MofCom said they will have a near auto licensing regime for open-source AI that allows more rapid release it seems? "To support the development of an open-source AI innovation ecosystem, automatic licensing and other facilitation measures will apply to technology exports involving open-source releases of AI technologies, Chinese companies building and operating their own computing centers overseas, overseas chip tape-outs and fabrication, joint R&D using overseas resources, and participation in international standard-setting—essentially allowing exports to proceed immediately upon filing." open on X
    2 likes · 0 reposts
  5. @natolambert research ·
    A safe assumption when people are all commenting on a short Chinese policy document is that they don't know what it means and tried to parse a partial translation from claude, without reading all the way through. There are maybe 5-10 on this site who i trust to comment.
    1 likes · 2 reposts
  6. @natolambert research ·
    A safe assumption when people are all commenting on a short Chinese policy document is that they don't know what it means and tried to parse a partial translation from claude, without reading all the way through. There are maybe 5-10 people on this site who i trust to comment.
    18 likes · 2 reposts
  7. @natolambert research ·
    In this case I normally would consider myself a "not knowing" but most of the takes have been so bad and extreme, that I'd upgrade myself to "knows better"
    1 likes · 0 reposts
  8. @natolambert research ·
    Where's your screenshot from?
    0 likes · 0 reposts

Reposted by @deanwball, @natolambert.