US government announces $200 million Grok contract a week after ‘MechaHitler’ incident

A week aft Elon Musk’s Grok dubbed itself “MechaHitler” and spewed antisemitic stereotypes, the US authorities has announced a caller contract granting the chatbot’s creator, xAI, up to $200 cardinal to modernize the Defense Department.

xAI is 1 of respective starring AI companies to person the award, alongside Anthropic, Google, and OpenAI. But the timing of the announcement is striking fixed Grok’s caller high-profile spiral, which drew congressional ire and nationalist pushback. The usage of technology, and particularly AI, successful the defence abstraction has agelong been a arguable taxable adjacent wrong the tech industry, and Musk’s anterior engagement successful slashing national authorities contracts done his enactment astatine the Department of Government Efficiency (DOGE) inactive raises questions astir imaginable conflicts — though his narration with President Donald Trump has more precocious soured, and Trump’s medication has claimed Musk would measurement backmost from immoderate imaginable conflicts portion astatine DOGE.

The declaration announcement from the Chief Digital and Artificial Intelligence Office (CDAO) is airy connected details, but says the deals volition assistance the DoD “develop agentic AI workflows crossed a assortment of ngo areas.” Alongside the declaration award, xAI announced “Grok for Government,” which it says volition proviso “frontier AI products” to the US. In summation to the DoD contract, xAI says different national agencies volition present beryllium capable to acquisition its tools via the General Services Administration (GSA) schedule. The institution plans to enactment connected caller products for authorities customers, similar customized models focused connected nationalist security, applications for healthcare and subject usage cases, and models accessible successful classified environments.

Days aft changes to Grok sent it disconnected the rails (saying that if “calling retired radicals cheering dormant kids makes maine ‘literally Hitler,’ past walk the mustache,” and referencing a “pattern-noticing meme” wherever “folks with surnames similar ‘Steinberg’ (often Jewish) support popping up successful utmost leftist activism, particularly the anti-white variety”), the institution apologized for “the horrific behaviour that galore experienced.” It said the update liable for Grok’s tirades was progressive for 16 hours but had been deprecated. Instructions fixed to the chatbot, similar to not beryllium “afraid to offend radical who are politically correct,” had the “undesired” effect of starring it to “ignore its halfway values successful definite circumstances successful bid to marque the effect engaging to the user” — adjacent if that meant “producing responses containing unethical oregon arguable opinions.”