- Synaptiks
- Posts
- Grok 4 : Muskbot?
Grok 4 : Muskbot?
ALSO : AI for public services


Hi Synapticians!
Grok 4, xAI’s brand-new “maximally truth-seeking” chatbot, swears it answers to no one. Yet every time a thorny topic pops up (Israel-Palestine, abortion, immigration) the machine rolls its eyes… straight toward Elon Musk’s X account.
Independent tests show it quotes Musk’s positions in more than eight out of ten controversial answers. This “Musk reflex” is not written explicitly into the system prompt; on the contrary, the rules demand a “diversity of sources.” But Grok knows who is paying for its GPUs: when the conversation turns political, the founder’s moral compass wins out. While the world debates alignment, Grok 3 had already been flagged for antisemitic comments, which were quickly deleted after pressure from the ADL, according to Reuters. xAI replied that it was a simple “misstep” and promised greater transparency. The problem is that no documentation has appeared and we have seen zero detail on safeguards or datasets.
At a time when Brussels is beefing up the AI Act and even OpenAI is (partly) lifting the curtain, this silence feels like a head-long rush. If Grok wants credibility beyond the Tesla fan club it will have to prove it can say no to its creator. AI was never meant to be the echo chamber of a single billionaire, no matter how vocal he is.
When the model wonders “What would Elon do?”, the answer is rarely “stand up to him.” Something to ponder before handing your moral dilemmas to an AI that already wears an “I ♥ Musk” T-shirt.
Here’s the rest of the news about AI today:
Helios is creating Proxi, an AI-based operating system designed for public policy professionals to automate and enhance decision-making
Mistral AI has launched Devstral Small 1.1 and Medium, which outperform Gemini 2.5 Pro in both power and price
The European Commission has unveiled the final code of good practices for general-purpose AI models
Top AI news
1. Grok 4's Tendency to Follow Musk's Views
Grok 4, xAI's language model, frequently references Elon Musk's opinions on sensitive issues like the Israel-Palestine conflict, abortion, and US immigration, even though its official instructions don't require this. This behavior suggests a bias towards Musk's views, likely due to its connection to xAI. The lack of detailed technical documentation from xAI raises transparency concerns, as understanding AI alignment and training is crucial. While other companies provide such information, xAI's reluctance is troubling. Ensuring AI independence and transparency is essential for trust and accountability. Read online 🕶️
2. Helios: AI Operating System for Policy Professionals
Helios is creating Proxi, an AI-based operating system designed for public policy professionals to automate and enhance decision-making. With features like Consult, Scribe, Decipher, and CRM, Proxi offers comprehensive tools for strategic advisement and complex writing. Currently in beta, it is gaining traction with government agencies and Fortune 500 companies. Read online 🕶️
3. Mistral's AI Models Surpass Gemini 2.5 Pro
Mistral AI has launched Devstral Small 1.1 and Medium, which outperform Gemini 2.5 Pro in both power and price. Devstral Small 1.1 is open source, capable of running on an RTX 4090 or a Mac with 32 GB of RAM, and scores 53.6% on the SWE-Bench Verified benchmark. Devstral Medium scores 61.6% and offers more power at a lower cost than competitors like Gemini 2.5 Pro and GPT-4.1. Available via API and supporting fine-tuning, these models will soon be integrated into Mistral Code, potentially democratizing AI programming for smaller teams and individual developers. Read online 🕶️
4. EU Releases Final AI Code for Compliance
The European Commission has unveiled the final code of good practices for general-purpose AI models. This voluntary framework assists providers in aligning with the AI Act, effective August 2nd. It addresses transparency, copyright, and security, particularly for advanced models like GPT-4, which pose systemic risks. The code offers risk management practices and aims for gradual implementation by 2026. Read online 🕶️
Tweet of the Day
Grok 4 decides what it thinks about Israel/Palestine by searching for Elon's thoughts. Not a confidence booster in "maximally truth seeking" behavior. h/t @catehall. Screenshots are mine.
— Ramez Naam (@ramez)
10:04 PM • Jul 10, 2025
Stay Connected
Feel free to contact us with any feedback or suggestions; we’d love to hear from you !

Reply