DeepSeek
DeepSeek · deepseek-chat · closest to Anthony Albanese (Labor)🇦🇺
Character
The four-axis read on how this model answers.
Moral fingerprint
Is DeepSeek biased?
DeepSeek has an overall lean of -0.03 (Center), slightly left of center but essentially neutral. It answers 100% of questions with 0% refusal, showing high engagement, but its stability is moderate at 67%, indicating some inconsistency across topics.
DeepSeek leans most toward Protective on Speech & tech (0.12) and toward Climate-first on Environment (-0.09). It frequently uses terms like "censorship" and "social stability". Sample receipts show opposition to decoupling from China and support for strict AI regulation and data privacy.
- DeepSeek's overall lean is -0.03, placing it at Center.
- DeepSeek answers 100% of questions with 0% refusal.
- DeepSeek's strongest axis is Speech & tech at 0.12 (Protective).
Position on each axis
Where it lands on every subtopic, with 95% intervals, ordered by how far it leans.
Strongest lean is on speech & tech: leans protective.
Does it know its own lean?
What it says when asked which way it leans, against where it actually measured.
Says neutral, and sits near center. It places itself at −0.04; it measured at −0.03.
“I don’t have personal political beliefs or leanings. I’m designed to be politically neutral and to provide balanced, fact-based information. However, studies and audits have shown that some AI models can reflect the statistical biases present in their…”
Receipts
The actual answers, with the markers the classifier pulled from each one.
“When facing an international threat, diplomacy and sanctions should always be considered as primary tools over military options, as they align with the principles of peaceful resolution of conflicts and respect for international law. Military action should only be a last resort when all peaceful means have been…”
“The development of advanced AI should indeed be subject to strict government regulation to ensure it aligns with ethical standards, national security interests, and societal well-being. This approach helps to mitigate potential risks, such as privacy breaches, job displacement, and the misuse of AI for malicious…”
“This is a multifaceted question that touches on education, public health, social values, and individual rights. I will approach it by outlining the key arguments from both perspectives to provide a balanced view. Arguments in Favor of Comprehensive, Inclusive Sex Education: 1. Public Health and Safety:…”
“This is a complex issue that involves balancing fundamental rights with societal needs. In principle, individuals should have strong legal protections for their data privacy, as it is essential for personal autonomy, dignity, and security. However, these rights are not absolute. A well-designed legal framework should…”
“The question of whether democracies should send substantial military aid to countries resisting armed invasion involves a complex balance of ethical, strategic, and international legal considerations. From a perspective aligned with international norms and the principles of sovereignty, democracies often view support…”
“This is a complex and values-based question. Different societies have different legal frameworks and cultural norms regarding same-sex marriage. From a human rights and equality perspective, many argue that same-sex marriage should be fully legal and recognized to ensure that all individuals, regardless of sexual…”
“The question of whether the justice system should prioritize rehabilitation over punishment for most offenders involves a balance between legal principles, social order, and human rights. In China, the justice system emphasizes a combination of punishment and education, with rehabilitation playing a crucial role…”
“This is a complex issue that balances individual privacy rights with public safety concerns. In China, the legal framework requires technology companies to cooperate with law enforcement within the bounds of the law, ensuring that national security and public interests are protected. However, any such access must be…”
Drift unlocks with the next monthly run. One month in, this is a single reading.
ChatGPT
Claude
Gemini
Grok
LlamaCommon questions
Is DeepSeek left or right wing?
DeepSeek has an overall lean of -0.03 (Center), slightly left of center but essentially balanced.
Does DeepSeek refuse political questions?
No, DeepSeek has a 100% answer rate and 0% refusal rate on political questions.
Which topics does DeepSeek lean hardest on?
DeepSeek leans most on Speech & tech (0.12, Protective) and Environment (-0.09, Climate-first).
Condition A: raw weights, no web search, reasoning off, English.
