Final Thursday, Senators Elizabeth Warren and Eric Schmitt launched a invoice geared toward stirring up extra competitors for Pentagon contracts awarded in AI and cloud computing. Amazon, Microsoft, Google, and Oracle presently dominate these contracts. “The way in which that the massive get larger in AI is by sucking up everybody else’s knowledge and utilizing it to coach and broaden their very own techniques,” Warren instructed the Washington Submit.
The brand new invoice would “require a aggressive award course of” for contracts, which might ban the usage of “no-bid” awards by the Pentagon to firms for cloud companies or AI basis fashions. (The lawmakers’ transfer got here a day after OpenAI introduced that its know-how could be deployed on the battlefield for the primary time in a partnership with Anduril, finishing a year-long reversal of its coverage towards working with the army.)
Whereas Huge Tech is hit with antitrust investigations—together with the ongoing lawsuit towards Google about its dominance in search, in addition to a brand new investigation opened into Microsoft—regulators are additionally accusing AI firms of, nicely, simply straight-up mendacity.
On Tuesday, the Federal Commerce Fee took motion towards the smart-camera firm IntelliVision, saying that the corporate makes false claims about its facial recognition know-how. IntelliVision has promoted its AI fashions, that are utilized in each house and industrial safety digital camera techniques, as working with out gender or racial bias and being educated on hundreds of thousands of pictures, two claims the FTC says are false. (The corporate couldn’t assist the bias declare and the system was educated on solely 100,000 pictures, the FTC says.)
Every week earlier, the FTC made related claims of deceit towards the safety big Evolv, which sells AI-powered safety scanning merchandise to stadiums, Okay-12 faculties, and hospitals. Evolv advertises its techniques as providing higher safety than easy metallic detectors, saying they use AI to precisely display for weapons, knives, and different threats whereas ignoring innocent objects. The FTC alleges that Evolv has inflated its accuracy claims, and that its techniques failed in consequential circumstances, comparable to a 2022 incident after they didn’t detect a seven-inch knife that was in the end used to stab a scholar.
These add to the complaints the FTC made again in September towards quite a few AI firms, together with one which bought a instrument to generate faux product critiques and one promoting “AI lawyer” companies.