AI & Astrology: An Ethical Framework

This framework exists because someone should write it before someone has to.

The application of artificial intelligence to astrological analysis is not a hypothetical. It is happening now. Consumer platforms generate daily horoscopes using large language models. Apps like Co-Star serve millions of users with algorithmically produced natal chart interpretations. Beaufort Intelligence uses AI within strict architectural constraints to produce structured psychological analysis from natal chart data. The question is no longer whether AI will be applied to astrology. The question is whether anyone will build the ethical constraints before the technology outpaces them.

Anthropic, the AI laboratory, published its Responsible Scaling Policy before its most capable models were deployed, arguing that safety frameworks must be established proactively because reactive regulation always arrives too late.^[1] The same logic applies here. If computational astrology matures to a point where its analytical outputs carry predictive weight, the implications for privacy, consent, discrimination, and autonomy are significant. Building the ethical constraints after that point is reached means building them after damage has already been done.

The history of analytical tools being repurposed as instruments of discrimination is not speculative. It is documented, repeated, and instructive.

IQ tests were designed for educational assessment. Lewis Terman adapted the Binet-Simon scale at Stanford and promoted its use for identifying children who needed additional support. Within a decade, the same instrument was being used to justify forced sterilisation programmes across the United States, classify racial groups as intellectually inferior, and restrict immigration from Southern and Eastern Europe. Terman himself was a member of the Human Betterment Foundation and the American Eugenics Society.^[2] The tool did not change. Its application did.

Personality assessments followed the same trajectory. The Myers-Briggs Type Indicator and the Big Five were developed for self-understanding and clinical insight. They are now used by employers to screen job candidates, a practice that the American Civil Liberties Union has challenged in formal complaints against hiring technology vendors, arguing that these assessments produce discriminatory outcomes that disproportionately exclude qualified candidates from protected groups.^[3] In 1971, the U.S. Supreme Court ruled in Griggs v. Duke Power Co. that employment tests with discriminatory impact violate civil rights law regardless of whether the employer intended to discriminate.^[4] Five decades later, the same pattern is being repeated with AI-driven assessments.

Credit scoring algorithms, trained on historically biased data, systematically disadvantage minority borrowers. Research from the Stanford Institute for Human-Centered Artificial Intelligence has shown that credit scores are 5 to 10 percent less accurate for minorities and low-income borrowers, not because of algorithmic bias per se, but because the data these algorithms are trained on reflects decades of structural inequality.^[5]

Genetic data offers the closest analogy to birth data. Before the Genetic Information Nondiscrimination Act was passed in the United States in 2008, employers and insurers could use genetic test results to deny coverage or employment.^[6] It took thirteen years of legislative effort, from the first bill introduced in 1995 to the law's enactment, to establish basic protections. And GINA still does not cover life insurance, disability insurance, or long-term care. Genetic data, like birth data, is immutable. The person cannot change it. The potential for discriminatory application existed from the moment the data became analytically useful.

The pattern across all four cases is identical. A tool is built for understanding. It is repurposed for classification. Classification becomes discrimination. Regulation arrives years or decades after the harm. The people who built the tool did not intend the harm, but the absence of proactive constraints made it inevitable.

Birth data occupies a unique position in the landscape of personal information. A person's date, time, and place of birth cannot be changed. Unlike a password, it cannot be reset. Unlike an address, it cannot be relocated. Unlike a name, it cannot be amended by deed poll. It is fixed at the moment of birth and remains constant for the duration of a person's life.

Under GDPR, date of birth is classified as personal data, and location data is explicitly covered.^[7] The full triad of birth date, birth time, and birth location does not currently fall within GDPR's Article 9 special categories (which cover genetic data, biometric data, health data, and religious or philosophical beliefs). But the combination of these three data points enables the generation of a complete natal chart, from which psychological and behavioural analysis can be derived. If that analysis becomes demonstrably accurate, the data that generates it becomes sensitive in practice, whether or not current regulation classifies it as such.

Consider, briefly, what becomes possible if computational astrology produces outputs with predictive validity. Insurance companies could request birth data alongside medical history, screening for natal configurations associated with risk-taking behaviour or health vulnerabilities. Employers could filter candidates by natal chart compatibility with company culture or team composition. Dating applications could sort potential matches by synastry scores. Governments have already used astrology for strategic planning: Nancy Reagan consulted astrologer Joan Quigley to influence President Reagan's schedule throughout his presidency, a practice documented in Quigley's own memoir and confirmed by Reagan's Chief of Staff, Donald Regan.^[8]

These scenarios are not science fiction. They are the logical extension of the same trajectory that transformed IQ tests, personality assessments, credit scores, and genetic data from tools of understanding into instruments of gatekeeping. The only variable is whether the constraints are built before or after.

The scenarios described above are institutional. They concern how organisations might misuse astrological data. But there is a deeper question that most ethical frameworks in this space have not confronted, because confronting it requires taking the subject matter seriously enough to follow the logic to its conclusion.

If computational astrology, augmented by artificial intelligence with sufficient pattern recognition and synthesis capability, were to produce outputs that are reliably and demonstrably accurate, the consequences extend far beyond data misuse. The entire materialist-rationalist framework that underpins Western science, medicine, law, and economics would have a hole in it. Not a small one. If planetary transits to natal positions can reliably predict specific psychological states or life events, that implies some form of non-local, non-causal connection between celestial mechanics and individual human experience. That is not an adjustment to the existing paradigm. It is a replacement of it.

The response would likely arrive in stages. Initial denial and debunking attempts from the scientific establishment, following the pattern documented in the accompanying literature review. Then, once the statistical evidence became undeniable, a fracture: one camp attempting to absorb the findings into physics through field theory or information-theoretic explanations, another interpreting them as vindication of pre-Enlightenment cosmology. Religious institutions would split on whether the findings confirm or threaten their own frameworks. The epistemological ground would shift beneath every discipline that depends on a mechanistic model of causation.

The practical consequences arrive faster than the philosophical ones, and they are darker. The commodification would be immediate and ruthless. Insurance companies would want birth charts. Employers would screen by natal placements. Courts might consider transit data as mitigating circumstances. Dating platforms would filter by synastry. Military and intelligence agencies would apply mundane astrology to strategic planning. Every institution that currently uses data to predict and sort human behaviour would have access to a new dataset that the individual cannot change, cannot opt out of, and may not even know is being used. The psychological depth that makes reflective astrological analysis valuable would be stripped out entirely, reducing a framework for self-understanding to a deterministic sorting tool.

There is also the question of what reliable predictive accuracy would do to the people receiving the analysis. If difficult transits can be identified in advance with genuine precision, a proportion of recipients would experience anticipatory fatalism. The capacity to see difficulty coming does not, for most people, produce equanimity. It produces anxiety. For individuals already experiencing psychological vulnerability, the delivery of accurate but distressing predictive information without adequate contextualisation and safeguarding could cause real harm. Any system operating in this space has an obligation to build protections against this, regardless of its current level of accuracy.

And the power asymmetry would be significant. Whoever controls the most capable interpretation models would control a form of predictive intelligence that makes every other forecasting methodology look primitive. Financial markets, geopolitical strategy, resource allocation. The concentration of that capability in any single entity, whether a corporation, a government, or an individual, would represent an asymmetry of information without historical precedent.

None of this is certain. The question of whether astrological analysis can produce outputs with genuine predictive validity remains open, and Beaufort Intelligence's research programme is designed to investigate it with methodological rigour. But the ethical framework cannot wait for the answer. If the constraints are built only after demonstrable accuracy is achieved, they will be built after the insurance companies, the employers, and the intelligence agencies have already begun using the data. The history of every analytical technology described in this paper confirms that reactive regulation arrives too late. The constraints must exist before they are needed.

The following principles govern Beaufort Intelligence's approach to applying artificial intelligence to astrological analysis. They are published openly and are intended to be adopted, adapted, or challenged by anyone building in this space.

These principles do not exist in a regulatory vacuum. The European Union's Artificial Intelligence Act, which entered into force in 2024, classifies AI systems that affect fundamental rights and psychological state as high-risk under Article 6 and Annex III.^[9] A system that generates psychological analysis from personal data and delivers it directly to the individual it describes falls squarely within this classification. High-risk systems under the EU AI Act are subject to enhanced requirements for transparency, data governance, human oversight, and risk management, requirements that align with the principles outlined above.

The American Psychological Association published guidance in 2024 on the ethical use of AI in professional practice, emphasising that AI tools used in psychological contexts must maintain the same ethical standards as human practitioners, including informed consent, competence boundaries, and avoidance of harm.^[10] Whether astrological analysis constitutes "psychological practice" is a definitional question that current regulation has not addressed. The prudent position is to operate as though it does.

The gap in existing regulation is not in the principles but in their application. GDPR covers birth data as personal data but does not classify it as sensitive. The EU AI Act addresses high-risk psychological systems but has not considered astrological analysis as a use case. APA guidance covers AI in professional practice but does not extend to non-clinical analytical tools. In every case, the regulatory framework has been built around existing technologies and existing categories of harm. Computational astrology sits in the space between those categories, which is precisely why a proactive framework is necessary.

The difference between building powerful tools and building responsible ones is not capability. It is constraint. Every analytical technology that has caused significant harm did so not because its creators were malicious but because the constraints were built after the applications, not before them. IQ tests existed for decades before the eugenics programmes they enabled were dismantled. Personality assessments have been used in discriminatory hiring for years while legal challenges are still working through the courts. Credit scoring algorithms continue to disadvantage the populations they were never designed to serve while regulators debate definitions.

Computational astrology is in its earliest stages. The question of whether natal chart analysis produces outputs with genuine predictive validity remains open and is being investigated through Beaufort Intelligence's formal research programme. But the ethical constraints should not depend on the answer to that question. If the analysis is accurate, the constraints protect people from its misuse. If it is not accurate, the constraints protect people from being harmed by false confidence in its outputs. In either case, the framework is necessary.

This document is published openly. It will be updated as the technology develops, as regulatory frameworks evolve, and as the research programme produces data that informs the ethical implications of this work. It is not a final statement. It is a starting position, established on record, before it is urgently needed.