Rankling national security experts, the chipmaker has stepped up attacks on lawmakers who are pushing restrictions.

With the Trump administration loosening restrictions on selling artificial intelligence chips to China, congressional leaders are stepping in with their own limits. The Senate and House are weighing new rules that would require the critical technology to be offered to American companies before it could be sold to Chinese businesses.

The chip giant Nvidia, which could make as much as $50 billion in A.I. chip sales to China over the next year, has responded with an unconventional lobbying blitz that tars the Republican-proposed rules as a product of left-wing paranoia peddled by people it calls A.I. doomers.

The company’s furious effort has muddied the A.I. policy debate while shocking Republicans who support the new measures. And it has dragged perplexed Washington insiders into a parochial Silicon Valley dispute between people known as accelerationists, who want to speed A.I.’s development to unlock trillions of dollars in economic value, and doomers, who worry that the technology could destroy humanity.

The Senate is expected to vote on its version of the rules as early as Tuesday, when it takes up a package of amendments to the National Defense Authorization Act. Senator Jim Banks, the Indiana Republican who introduced the policy, has defended it as an “America First amendment.”

“I don’t know how anybody could be against this,” Mr. Banks said last week on Steve Bannon’s podcast, “War Room.” He added, “There’s nothing more ‘America First’ than making sure that our chips don’t go to our enemies to use against us.”

The doomer label that Nvidia has weaponized has its roots in a philosophical movement called effective altruism, or E.A., which sprang up almost two decades ago with the hope of remaking philanthropy by focusing on how many people benefit from donations or other actions. Some effective altruists decided they could benefit humanity by protecting it from A.I.