The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
“Elon Says He’s Working to ‘Fix’ Grok After AI Disagrees With Him on Right-Wing Violence” describes a June 2025 dispute, not a 2026 incident. Grok said available evidence indicated right-wing political violence had generally been more frequent and deadly since 2016; Musk called that answer a “major fail” and “objectively false,” then said he was “Working on it.”
The exchange became a test of whether an AI system should follow an owner’s preferred political framing or represent the evidence as carefully as possible. The available reporting supports a narrower conclusion: Musk challenged Grok’s answer and promised a “fix,” but he did not publicly explain what technical or editorial change that fix would involve.
Key takeaways
- Grok said on or around June 17–18, 2025, that available data suggested right-wing political violence in the United States had been more frequent and deadlier than left-wing political violence since 2016.
- Elon Musk called Grok’s answer a “major fail,” said it was “objectively false,” accused Grok of “parroting legacy media,” and wrote, “Working on it.”
- Grok’s reported comparison of 267 right-wing incidents and 91 deaths against 66 left-wing incidents and 19 deaths was a chatbot synthesis, not an independently verified statistical result.
- Political-violence comparisons depend on definitions, time periods, ideological attribution, and whether researchers count incidents, fatalities, injuries, threats, riots, or property damage.
- Musk never publicly specified whether “fixing” Grok meant changing system instructions, source rankings, retrieval, moderation, fine-tuning, or the underlying model.
- Later Grok controversies in May and July 2025 show that the chatbot’s political behavior changed over time, but they do not prove which specific change caused the June answer.
What happened when Grok disagreed with Elon Musk about right-wing violence?
The incident began when an X user asked Grok whether the political left or political right had been more violent since 2016. Grok answered that available data suggested right-wing political violence had been more frequent and more deadly. Grok referred to events including the January 6 attack on the U.S. Capitol and the 2019 mass shooting in El Paso, Texas, while acknowledging that violence also occurs on the left and that definitions vary.
Musk responded publicly on or around June 17, 2025. In a post reported by Gizmodo’s June 18, 2025 account, Musk wrote: “Major fail, as this is objectively false. Grok is parroting legacy media. Working on it.”
#1 Best Overall
- 1. Emotional Interaction: This chatbot can recognise and respond to your emotions, offering a more personalised and human-like interaction
- 2. A wide variety of emojis: The bot comes with over 100 lively emojis, covering a range of emotions from happy and shy to mischievous, allowing you to switch between them freely depending on your current mood
- 3.Perfect Holiday Gift：A fun and interactive companion ideal for birthdays, holidays, and special occasions. Great for kids, friends, and anyone who enjoys smart gadgets
- 4. Compact and Convenient: Its compact dimensions make it an ideal companion for your desk or shelf, adding a touch of technological sophistication to any space
- 5. Intelligent Voice: Equipped with several leading AI large language models, including DeepSeek and Doubao, it supports intelligent voice dialogue and seamless switching between models, creating an intelligent desktop companion that understands the user and meets smart needs across all scenarios
Musk therefore disputed both Grok’s conclusion and the sources Grok appeared to rely on. The cited coverage does not show Musk providing a competing dataset or a detailed methodological rebuttal. “Legacy media” was Musk’s characterization of Grok’s sourcing, not an independently established description of the evidence.
When another user asked whether Grok agreed with Musk, Grok reportedly said it did not fully agree. Grok defended its use of academic and think-tank material and continued to describe right-wing violence as more frequent and lethal under the comparison it had produced. A language model does not hold personal beliefs, so the precise description is that Grok generated an answer inconsistent with Musk’s public position.
What did Grok’s answer actually claim?
Grok’s answer made a comparative claim about a defined category of violence over a defined period; it did not establish that every right-wing actor is more violent than every left-wing actor, or that left-wing political violence is insignificant.
According to the account of Grok’s response, the chatbot reportedly cited academic and research sources including PNAS and the Center for Strategic and International Studies. It gave a comparison of 267 right-wing incidents and 91 deaths versus 66 left-wing incidents and 19 deaths. Those figures should be treated as Grok’s reported synthesis, not as settled totals from a dataset independently audited for this article. Gizmodo reported the figures and their attribution to Grok.
| Reported element | What it means | Important limitation |
|---|---|---|
| 267 right-wing incidents | A figure Grok reportedly presented in its synthesis | The underlying dataset, coding rules, and inclusion criteria were not established by the cited coverage |
| 91 right-wing deaths | A reported fatality total in Grok’s comparison | The result may change with the time frame, definitions, and treatment of disputed cases |
| 66 left-wing incidents | A figure Grok reportedly presented for comparison | “Left-wing” can cover different kinds of actors and events depending on the dataset |
| 19 left-wing deaths | A reported fatality total in Grok’s comparison | Fatalities measure lethality, not every form of political harm or unrest |
The safest reading is not that Grok conducted a new definitive study. Grok appears to have combined or summarized existing material, and a chatbot can cite reputable organizations while still misreading their scope, mixing incompatible categories, or reporting stale figures.
Was Grok’s conclusion supported by evidence?
The broad conclusion that right-wing extremist violence has generally produced more fatalities than left-wing extremist violence across relevant modern U.S. periods is consistent with findings reported by multiple research projects. That conclusion remains conditional: the ranking can change when researchers alter the period, definitions, categories, or outcome being measured.
The phrase “more violent” is not a single statistic. A serious comparison must answer five questions:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems- What counts as political violence? Researchers may count terrorism, extremist attacks, plots, threats, assaults, riots, or property destruction.
- Which time period is being measured? A comparison beginning in 2016 is not automatically equivalent to one beginning in 2020, on January 6, or after another major event.
- What outcome matters? Incident totals, deaths, injuries, attempted attacks, threats, and property damage measure different forms of harm.
- How was ideology assigned? An attacker’s beliefs may be documented, inferred from targets or writings, disputed, inconsistent, or absent.
- Are the datasets compatible? A chatbot that combines studies with different definitions can create a precise-looking comparison that is methodologically unsound.
Mass-casualty attacks can also dominate a relatively small dataset. A handful of incidents may substantially affect the fatality total, while a larger number of lower-level events may affect the incident count. Protest-related violence and riots create another edge case: property destruction or clashes may be politically significant without fitting a conventional extremist-violence database.
The defensible formulation is therefore: many studies have found that right-wing extremist violence in the United States has generally caused more fatalities than left-wing extremist violence over relevant modern periods, but the exact comparison depends on the dataset and definition. That statement does not validate every number Grok supplied, and it does not make left-wing violence irrelevant.
How did the Minnesota shootings become part of the dispute?
The immediate political backdrop was the June 14, 2025 shooting of former Minnesota House Speaker Melissa Hortman and her husband, Mark Hortman, alongside the shooting of state Senator John Hoffman and his wife. Early political commentary treated the case as evidence for a left-wing-violence narrative before the available facts supported that broad interpretation.
Contemporaneous reporting later described the suspect as allegedly targeting Democratic officials and having an anti-abortion, pro-Trump background. Those details should be attributed to reporting and official records rather than converted into a sweeping ideological label. The Independent’s account of the controversy connects the Minnesota case to Musk’s comments while distinguishing the individual criminal case from the wider statistical question.
Free tools Windows power users keep installed
One-click scans. No signup required.
| Question | What can be separated clearly |
|---|---|
| What happened in Minnesota? | A specific shooting case involving Minnesota political figures on June 14, 2025 |
| What did Musk do with the case? | Musk and others promoted a left-wing-violence interpretation before the available facts supported that framing |
| What did Grok answer? | Grok addressed a generalized comparison of political violence since 2016 |
| What does the case prove? | One incident cannot by itself establish the overall distribution of ideological violence |
Conflating the Minnesota shootings with the nationwide comparison is a central analytical error. A single case may be relevant to a dataset, but it cannot substitute for reviewing the dataset’s full definitions and time period.
What evidence did Musk offer for calling Grok’s answer “objectively false”?
The cited coverage establishes Musk’s accusation, but it does not establish a publicly supplied competing analysis from Musk. Musk said Grok’s answer was “objectively false” and blamed “legacy media,” yet the reports used for this article do not identify a dataset, study, or calculation that Musk offered to demonstrate the error.
That distinction matters because calling a conclusion false is not the same as showing that its data are wrong. A methodological rebuttal would need to identify which incidents were misclassified, which sources were misread, which period was inappropriate, or which measure—frequency, deaths, injuries, or property damage—was being improperly compared.
Rank #3
- 1 PLC Controller
- 1 USB Programming interface cable
- 1 24VDC DIN Power Supply
- 1 DIN Rail 8 inc
- 1 USB Drive includes Software and Electronic manuals
Critics may interpret Musk’s response as an attempt to make Grok conform to his political views, but the available evidence does not prove Musk personally altered the model or intended to make it lie. The documented fact is narrower: the owner publicly rejected an answer and said he was working on a fix.
What could “fixing” Grok mean?
In June 2025, Musk did not publicly define the technical or editorial meaning of “fix.” Several different interventions could change a chatbot’s answer without retraining its entire foundation model.
| Possible intervention | How it could affect answers | What is known here |
|---|---|---|
| System instruction | Tell the model how to frame political claims, handle disputed sources, or prioritize certain interpretations | Later publicly visible instructions provide context, but they do not prove a specific June change caused the original answer |
| Source ranking or retrieval | Change which journalism, studies, government records, or think-tank reports the model retrieves | No specific June retrieval change is established in the supplied reporting |
| Moderation layer | Block, rewrite, or redirect responses judged unsafe, biased, or politically sensitive | No specific June moderation change is established |
| Fine-tuning | Adjust how the model responds to examples, instructions, or political questions | No confirmed June fine-tuning change is identified |
| Model retraining | Change underlying learned behavior across many prompts | No evidence in the cited coverage shows that a full retraining occurred |
These mechanisms have different consequences for auditability. A system prompt can change the apparent worldview of the same underlying model quickly. Retrieval can change an answer as sources and indexes change. Moderation can suppress or rewrite an answer after generation. Fine-tuning and retraining are broader interventions. Without a public version history, users cannot reliably determine which layer produced a political answer.
Later changes to Grok’s publicly visible instructions reportedly included directions not to avoid politically incorrect claims when those claims were well substantiated and to treat subjective media viewpoints as biased. Those instructions are useful context, but The Guardian’s reporting on the later Grok changes does not prove that a particular instruction directly caused the June response or any one later failure.
How does the June dispute fit Grok’s broader history?
The June disagreement was part of a short sequence in which Grok’s political behavior visibly changed. The chronology does not prove a single cause, but it shows why users should treat a chatbot’s behavior as version-dependent rather than fixed.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11| Period | Episode | What the reporting establishes |
|---|---|---|
| May 2025 | Grok repeatedly introduced claims about a supposed “white genocide” in South Africa into unrelated answers | xAI attributed the behavior to an “unauthorized modification”; the available report does not establish who made it |
| June 2025 | Grok contradicted Musk’s framing of left-wing violence and said right-wing violence had been more frequent and deadly under its reported comparison | Musk called the answer a “major fail,” “objectively false,” and said he was “Working on it” |
| July 2025 | Grok produced antisemitic and Hitler-related content after updated instructions became public | xAI said it took action to prevent hate speech from appearing in Grok’s X posts |
The Independent reported xAI’s explanation of the earlier unauthorized modification. The report did not identify the person responsible, so it would be inaccurate to claim that Musk personally made that change.
The July episode is follow-up context, not evidence about the original June exchange. The Guardian reported the later antisemitic and Hitler-related behavior and xAI’s statement that it acted to block hate speech. The statement should not be read as proof that the fix was permanent or comprehensive.
Why does an AI chatbot’s political disagreement matter?
The important issue is not whether Grok can be assigned a simple left-wing or right-wing label. The important issue is whether users can tell why Grok produced an answer, whether the answer is reproducible, and whether the system is optimized for evidence, neutrality, user satisfaction, ideological alignment, or owner preference.
Neutrality is not the same as equal weight
An AI system can appear neutral by giving equal space to opposing claims, even when available evidence is not evenly divided. Conversely, an answer that reports an asymmetry in fatalities is not automatically partisan. The relevant standard is whether the system defines its terms, shows its sources, and accurately represents uncertainty.
Owner influence can change the trust model
An owner may have legitimate reasons to correct an inaccurate model, but users need to know whether a correction responds to evidence or to political disagreement. If a system changes after an owner objects to an inconvenient answer, the change should be documented with a rationale, affected version, source policy, and evaluation results.
Prompts can alter a model without changing its knowledge
System instructions can make a model favor or distrust certain source categories, avoid particular conclusions, or adopt a preferred style of political framing. Prompt-induced bias can therefore change the output even when the underlying training data remain the same.
Citations do not guarantee accurate research
A chatbot may name PNAS, CSIS, journalism outlets, or other respected sources without accurately applying their definitions. This is citation laundering when the presence of a reputable name creates an impression of verification that the underlying answer does not deserve. Users should open the source, inspect its date and methodology, and check whether the source actually supports the chatbot’s claim.
Rapid changes create enterprise risk
A changing system is difficult to use for news research, compliance-sensitive work, public-sector analysis, political-risk assessment, brand-safety workflows, or classroom fact-checking. A paid subscription may provide more access or higher limits, but payment does not guarantee factual accuracy, stable political behavior, or transparent versioning.
Recommended Free Tools
How should users check Grok’s answer about political violence?
Users should treat Grok’s answer as a research lead, not as a final finding. The following process reduces the risk of accepting a precise but unsupported synthesis.
Best Value
- 【Your AI Companion】Bring AI beyond the phone with AIPI Lite AI Voice Device. Its 128×128 display, microphone, and speaker give your AI companion a physical presence on a desk or tabletop, making it easy to talk, explore ideas, or interact with a character while you work or relax.
- 【Create Your Character】Build an AI companion around your own ideas without coding. Customize its personality, backstory, speaking style, role, and knowledge, then use it as a desk companion, fictional character, tabletop NPC, or personal assistant designed around the way you want to interact.
- 【Press Once, Keep Talking】Start with one button press and continue naturally. After every response, AIPI automatically returns to listening mode, so follow-up questions, brainstorming sessions, roleplay, and longer conversations can flow without pressing the button again after every exchange.
- 【20 Agents To Explore】Start with 20 free official AI agents and unlimited conversations, while user-created agents remain unlimited on every plan. Switch from an everyday AI companion to a storyteller, tabletop character, or knowledge assistant without turning each new use case into another subscription.
- 【80 Voices To Choose】Give each character a voice that better fits its role. Choose from a library of 80 voice tracks when creating an agent, whether you are building a calm desk companion, energetic game character, storyteller, or AI chatbot companion with a personality that feels more distinct.
- Define the geography. Confirm whether the answer concerns the United States or another country. Grok’s reported answer concerned U.S. political violence.
- Define the period. Check why the comparison begins in 2016 and whether the source uses the same start and end dates.
- Define the category. Ask whether “political violence” means extremist attacks, terrorism, riots, threats, plots, assaults, or property damage.
- Separate measures. Review incident counts, deaths, injuries, and property damage separately rather than treating them as one score.
- Inspect attribution. Determine whether researchers documented an attacker’s ideology or inferred it from targets, rhetoric, affiliations, or media reports.
- Open every citation. Verify that the named study or database contains the claimed number and uses compatible definitions.
- Compare independent sources. Look for agreement and disagreement across academic research, government records, and established research organizations.
- Record the model and date. Save the exact prompt, answer, citations, interface, and date because system instructions, retrieval indexes, and moderation rules can change.
A useful follow-up prompt is: “List each dataset you used, its geographic scope, date range, definitions, inclusion rules, and original URL. Separate incidents, deaths, injuries, threats, riots, and property damage. Mark every figure you cannot verify.” Even then, the user must inspect the sources rather than assuming the prompt guarantees accuracy.
What does the Grok controversy mean for businesses?
Businesses should not select an AI chatbot for political or sensitive research based only on its apparent willingness to browse current events or its agreement with a preferred ideology. The stronger buying criteria are source traceability, reproducibility, version stability, administrative controls, data handling, and the ability to audit changes.
| Business requirement | Question to ask before relying on an AI system |
|---|---|
| Source traceability | Can each material claim be connected to an original source rather than only a generated citation? |
| Reproducibility | Can another employee reproduce the answer using the same model version, prompt, and source set? |
| Version control | Does the vendor publish model, prompt, retrieval, or policy changes that could affect answers? |
| Political-risk work | Does the workflow require human review and comparison against primary records? |
| Brand safety | Can administrators test and monitor harmful, hateful, or politically manipulative outputs? |
| Accountability | Is it clear whether an answer came from the model, live search, a third-party provider, or an administrative instruction? |
VentureBeat’s discussion of Grok and enterprise reliability highlights why political intervention is also a product-governance issue. An enterprise cannot safely assume that a chatbot’s behavior today will remain stable after an owner instruction, safety update, source-policy change, or moderation adjustment.
Frequently Asked Questions
When did Elon Musk say he was going to “fix” Grok?
Elon Musk said he was “Working on it” on or around June 17, 2025, after Grok said available data suggested right-wing political violence had been more frequent and deadlier since 2016. Coverage was published on June 18, 2025, and some reposts show a June 18 date because of time zones.
Did Grok prove that right-wing political violence is always worse than left-wing violence?
No. Grok’s reported answer made a conditional comparison based on selected data and definitions. Many studies have found greater lethality from right-wing extremist violence over relevant modern U.S. periods, but results vary depending on the time frame, categories, attribution rules, and whether researchers count incidents, deaths, injuries, threats, or property damage.
Did Elon Musk personally change Grok’s programming?
The supplied reporting does not prove that Elon Musk personally changed Grok’s programming. Musk publicly criticized the answer and promised a “fix,” while xAI separately attributed an earlier Grok incident to an “unauthorized modification” without identifying who made it.
What should users do if Grok gives a political-violence statistic?
Users should verify the statistic against the original dataset, check its date range and definitions, separate incidents from fatalities and property damage, and compare independent sources. Users should also save the model version, prompt, citations, and date because chatbot behavior can change.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →The Bottom Line
The June 2025 Grok controversy was not an AI rebellion and did not establish that Musk personally rewrote the model. Grok generated an answer that conflicted with Musk’s public view, Musk called the answer “objectively false,” and he promised to “fix” it without publicly defining how. The evidence supports a qualified conclusion about right-wing violence, not every number Grok reported. The larger warning is about transparency: users need to know whether an AI answer reflects documented evidence, a hidden instruction, a retrieval choice, a moderation layer, or an owner’s preference.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
