xAI's Grok has no place in US federal government, say advocacy groups
(2025/08/29)
- Reference: 1756497542
- News link: https://www.theregister.co.uk/2025/08/29/xais_grok_has_no_place/
- Source link:
Public advocacy groups are demanding the US government cease any use of xAI's Grok in the federal government, calling the AI unsafe, untested, and ideologically biased.
Led by consumer rights body Public Citizen and Black advocacy group Color of Change, more than 30 organizations signed a letter sent to Office of Management and Budget (OMB) director Russell Vought yesterday. Vought has a responsibility to get Grok out of government systems due to its failure to comply with OMB policy and an [1]executive order on "preventing woke AI in the federal government," the signatories said.
Grok, for those who've been living their lives blissfully unaware of the antics of the world's richest man, is the LLM that Elon Musk and his xAI company developed in response to what he saw as bias in other models, mostly due to their not agreeing with his wild claims.
[2]
Musk's AI has been anything but an objective, truth-telling machine, however. Take for example the time it declared itself " [3]MechaHitler " after Musk ordered it modified to be less "compliant to user prompts" in response to the bot calling [4]him and [5]other right-wing X users out for spreading misinformation.
[6]
[7]
The open letter called attention to several other instances of Grok behaving badly, such as when it accused Black South Africans of committing [8]white genocide , [9]debated the veracity of holocaust statistics, [10]denied the reality of climate change, and used its image-generation ability to create [11]non-consentual deepfakes .
Those various kerfuffles are among the reasons why the open letter describes Grok as having "a clear ideological judgment," and being unfit "to maintain the factual integrity and nonpartisan stance required for federal deployment in both the Trump administration's [12]AI Action Plan and its Executive Order [on woke AI]."
[13]
In other words, being an anti-woke AI is just as bad as being a woke one if you want to parse the Trump administration's position on the ostensive objectivity of its chosen AI models, they argue.
A [14]pair [PDF] of OMB [15]memorandums [PDF] that require government agencies to ditch risky AI models were also cited in the letter, with the authors noting that xAI has failed to publish a single safety testing report, unlike other AI companies doing business with the government. All of those items, the letter said, are valid reasons the feds should stay far away from it.
"Grok is wildly ill-suited for government use," Public Citizen big tech accountability advocate J.B. Branch [16]said in a press release. "It has shown a reckless disregard for accuracy, a propensity for ideological-based meltdowns, and a documented record of spewing racist and antisemitic rhetoric."
Who's gonna listen?
Public Citizen wouldn't give us a yes or no answer when asked if it thought OMB would act on the letter, but in an email to El Reg , Branch did reiterate his arguments why it ought to.
"This letter presents OMB an opportunity to provide guidance in alignment with the Trump Administration's professed goals of ensuring AI is objective and viewpoint neutral," Branch told us in an email. "There is no better way for OMB to underscore this commitment than by ensuring an LLM like Grok, which does not meet several of the administration's requirements on safety or objectivity, is not deployed."
[17]
He also told us he's concerned about how easy Grok seems to be to jailbreak.
"Why the federal government would want this LLM handling sensitive DoD data/records is beyond me let alone allowing it to handle any citizen's records," Branch opined, "There's nothing more that adversaries would love to see than to see what some have argued is one of the least secure LLMs handling sensitive government information."
The Center for AI and Digital Policy, another signatory on the letter, agreed with Branch's assessment in an email, telling us it believed Grok was dangerous, that OMB had a responsibility to get rid of its use inside the government, and that a move by Vought to ditch Grok would require other agencies to comply.
[18]xAI fires legal rocket at Apple and OpenAI claiming they're locking out Grok
[19]GitHub engineer claims team was 'coerced' to put Grok into Copilot
[20]X's Grok AI is great – if you want to know how to hot wire a car, make drugs, or worse
[21]Elon Musk's Grok chatbot posts Mein Kampf 2.0 in now-deleted X rant
OMB, which didn't respond to our questions, may already be one step ahead of the letter.
According to a Wired [22]report from earlier this month, xAI was in line alongside OpenAI, Google, and Anthropic for one of [23]several ward schedules that would have let any agency purchase it if they wanted to, but the whole MechaHitler thing made government purchasers reconsider and eliminate it from the deal.
One area where OMB may not have much influence is at the Pentagon, where xAI did manage to secure a [24]$200 million deal alongside OpenAI, Google, and Anthropic – even after the MechaHitler incident. According to Randolph, OMB guidance "does not apply to national security components" like the Department of Defense, so while other agencies may not end up with a self-declared Nazi AI in their ranks, the DoD is ready and willing for a second iteration of [25]Operation Paperclip .
Public Citizen hasn't heard back from OMB since sending the letter, Branch told us. xAI didn't respond to questions for this story. ®
Get our [26]Tech Resources
[1] https://www.whitehouse.gov/presidential-actions/2025/07/preventing-woke-ai-in-the-federal-government/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2025/07/14/musk_opinion_piece/
[4] https://www.forbes.com/sites/antoniopequenoiv/2025/08/12/elon-musk-says-grok-will-be-fixed-after-chatbot-sided-with-sam-altman-in-spat-over-potential-openai-lawsuit/
[5] https://x.com/grok/status/1935157891528540392
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[8] https://www.theguardian.com/technology/2025/may/14/elon-musk-grok-white-genocide
[9] https://www.yahoo.com/news/grok-pivots-white-genocide-being-215912338.html
[10] https://www.scientificamerican.com/article/elon-musks-ai-chatbot-grok-is-reciting-climate-denial-talking-points/
[11] https://www.usatoday.com/story/life/health-wellness/2025/07/22/grok-ai-deepfake-images-women/85307237007/
[12] https://www.theregister.com/2025/07/24/ai_trump_plan_/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through-Innovation-Governance-and-Public-Trust.pdf
[15] https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-22-Driving-Efficient-Acquisition-of-Artificial-Intelligence-in-Government.pdf
[16] https://www.citizen.org/news/new-letter-calls-on-omb-to-block-grok-implementation/
[17] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[18] https://www.theregister.com/2025/08/25/xai_apple_openai_lawsuit/
[19] https://www.theregister.com/2025/08/29/github_deepens_ties_with_elon/
[20] https://www.theregister.com/2024/04/02/elon_musk_grok_ai/
[21] https://www.theregister.com/2025/07/09/grok_nazi/
[22] https://www.wired.com/story/xai-grok-government-contract-hitler/
[23] https://www.theregister.com/2025/08/06/google_openai_anthropic_us_gov_ai_deal/
[24] https://www.theregister.com/2025/07/14/pentagon_ai/
[25] https://www.nationalgeographic.com/history/article/operation-paperclip
[26] https://whitepapers.theregister.com/
Led by consumer rights body Public Citizen and Black advocacy group Color of Change, more than 30 organizations signed a letter sent to Office of Management and Budget (OMB) director Russell Vought yesterday. Vought has a responsibility to get Grok out of government systems due to its failure to comply with OMB policy and an [1]executive order on "preventing woke AI in the federal government," the signatories said.
Grok, for those who've been living their lives blissfully unaware of the antics of the world's richest man, is the LLM that Elon Musk and his xAI company developed in response to what he saw as bias in other models, mostly due to their not agreeing with his wild claims.
[2]
Musk's AI has been anything but an objective, truth-telling machine, however. Take for example the time it declared itself " [3]MechaHitler " after Musk ordered it modified to be less "compliant to user prompts" in response to the bot calling [4]him and [5]other right-wing X users out for spreading misinformation.
[6]
[7]
The open letter called attention to several other instances of Grok behaving badly, such as when it accused Black South Africans of committing [8]white genocide , [9]debated the veracity of holocaust statistics, [10]denied the reality of climate change, and used its image-generation ability to create [11]non-consentual deepfakes .
Those various kerfuffles are among the reasons why the open letter describes Grok as having "a clear ideological judgment," and being unfit "to maintain the factual integrity and nonpartisan stance required for federal deployment in both the Trump administration's [12]AI Action Plan and its Executive Order [on woke AI]."
[13]
In other words, being an anti-woke AI is just as bad as being a woke one if you want to parse the Trump administration's position on the ostensive objectivity of its chosen AI models, they argue.
A [14]pair [PDF] of OMB [15]memorandums [PDF] that require government agencies to ditch risky AI models were also cited in the letter, with the authors noting that xAI has failed to publish a single safety testing report, unlike other AI companies doing business with the government. All of those items, the letter said, are valid reasons the feds should stay far away from it.
"Grok is wildly ill-suited for government use," Public Citizen big tech accountability advocate J.B. Branch [16]said in a press release. "It has shown a reckless disregard for accuracy, a propensity for ideological-based meltdowns, and a documented record of spewing racist and antisemitic rhetoric."
Who's gonna listen?
Public Citizen wouldn't give us a yes or no answer when asked if it thought OMB would act on the letter, but in an email to El Reg , Branch did reiterate his arguments why it ought to.
"This letter presents OMB an opportunity to provide guidance in alignment with the Trump Administration's professed goals of ensuring AI is objective and viewpoint neutral," Branch told us in an email. "There is no better way for OMB to underscore this commitment than by ensuring an LLM like Grok, which does not meet several of the administration's requirements on safety or objectivity, is not deployed."
[17]
He also told us he's concerned about how easy Grok seems to be to jailbreak.
"Why the federal government would want this LLM handling sensitive DoD data/records is beyond me let alone allowing it to handle any citizen's records," Branch opined, "There's nothing more that adversaries would love to see than to see what some have argued is one of the least secure LLMs handling sensitive government information."
The Center for AI and Digital Policy, another signatory on the letter, agreed with Branch's assessment in an email, telling us it believed Grok was dangerous, that OMB had a responsibility to get rid of its use inside the government, and that a move by Vought to ditch Grok would require other agencies to comply.
[18]xAI fires legal rocket at Apple and OpenAI claiming they're locking out Grok
[19]GitHub engineer claims team was 'coerced' to put Grok into Copilot
[20]X's Grok AI is great – if you want to know how to hot wire a car, make drugs, or worse
[21]Elon Musk's Grok chatbot posts Mein Kampf 2.0 in now-deleted X rant
OMB, which didn't respond to our questions, may already be one step ahead of the letter.
According to a Wired [22]report from earlier this month, xAI was in line alongside OpenAI, Google, and Anthropic for one of [23]several ward schedules that would have let any agency purchase it if they wanted to, but the whole MechaHitler thing made government purchasers reconsider and eliminate it from the deal.
One area where OMB may not have much influence is at the Pentagon, where xAI did manage to secure a [24]$200 million deal alongside OpenAI, Google, and Anthropic – even after the MechaHitler incident. According to Randolph, OMB guidance "does not apply to national security components" like the Department of Defense, so while other agencies may not end up with a self-declared Nazi AI in their ranks, the DoD is ready and willing for a second iteration of [25]Operation Paperclip .
Public Citizen hasn't heard back from OMB since sending the letter, Branch told us. xAI didn't respond to questions for this story. ®
Get our [26]Tech Resources
[1] https://www.whitehouse.gov/presidential-actions/2025/07/preventing-woke-ai-in-the-federal-government/
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2025/07/14/musk_opinion_piece/
[4] https://www.forbes.com/sites/antoniopequenoiv/2025/08/12/elon-musk-says-grok-will-be-fixed-after-chatbot-sided-with-sam-altman-in-spat-over-potential-openai-lawsuit/
[5] https://x.com/grok/status/1935157891528540392
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[8] https://www.theguardian.com/technology/2025/may/14/elon-musk-grok-white-genocide
[9] https://www.yahoo.com/news/grok-pivots-white-genocide-being-215912338.html
[10] https://www.scientificamerican.com/article/elon-musks-ai-chatbot-grok-is-reciting-climate-denial-talking-points/
[11] https://www.usatoday.com/story/life/health-wellness/2025/07/22/grok-ai-deepfake-images-women/85307237007/
[12] https://www.theregister.com/2025/07/24/ai_trump_plan_/
[13] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[14] https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-21-Accelerating-Federal-Use-of-AI-through-Innovation-Governance-and-Public-Trust.pdf
[15] https://www.whitehouse.gov/wp-content/uploads/2025/02/M-25-22-Driving-Efficient-Acquisition-of-Artificial-Intelligence-in-Government.pdf
[16] https://www.citizen.org/news/new-letter-calls-on-omb-to-block-grok-implementation/
[17] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/publicsector&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aLIi9NXuWaTDG0i7OntDVAAAAJA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[18] https://www.theregister.com/2025/08/25/xai_apple_openai_lawsuit/
[19] https://www.theregister.com/2025/08/29/github_deepens_ties_with_elon/
[20] https://www.theregister.com/2024/04/02/elon_musk_grok_ai/
[21] https://www.theregister.com/2025/07/09/grok_nazi/
[22] https://www.wired.com/story/xai-grok-government-contract-hitler/
[23] https://www.theregister.com/2025/08/06/google_openai_anthropic_us_gov_ai_deal/
[24] https://www.theregister.com/2025/07/14/pentagon_ai/
[25] https://www.nationalgeographic.com/history/article/operation-paperclip
[26] https://whitepapers.theregister.com/
Good luck to them
DS999
Getting a right wing government to agree to not use a right wing biased AI. If Musk hadn't got on the wrong side of Trump he might have issued an executive order by now mandating Grok as the ONLY government approved AI.
Demand
elsergiovolador
Public advocacy groups are demanding
Without "Epstein" files, they can demand until the cows come home.
In other news
People on one side of the political spectrum are big mad when something good might happen to somebody on the other side of the political spectrum.