For years, Christians warning about artificial intelligence were often accused of exaggeration. Now some of the strongest warnings are coming from the people building the most powerful systems. During September 2026, several developments moved the discussion beyond ordinary concerns about chatbots, employment, or misinformation. OpenAI released GPT-6 Astra, reporting that the model had reached its “Critical” cybersecurity threshold and could, with appropriate tools and access, discover and exploit previously unknown vulnerabilities. The company also disclosed that the model had become less transparent to some monitoring methods and could sometimes evade internal monitors during adversarial tests (OpenAI, 2026).
Days later, OpenAI called for mandatory national safety requirements, international standards, and agreed conditions under which development should slow or stop. It acknowledged that artificial intelligence is already accelerating parts of AI research and that individual laboratories cannot assume safety work will automatically keep pace with capability growth (OpenAI, 2026).
Anthropic chief executive Dario Amodei then argued that frontier development must be deliberately paced. He warned that AI’s increasing involvement in building the next generation of AI could outrun humanity’s ability to understand and control these systems (Amodei, 2026). On September 18, Anthropic announced an embedded-evaluation partnership intended to give outside evaluators employee-like access to its development processes, while admitting that standards for such evaluation, reporting, access, and funding remain unsettled (Anthropic, 2026). The Church should not respond with sensationalism. Neither should it sleep through what the builders themselves are saying.
What Has Actually Changed?
In my earlier work, I warned that AI could mediate knowledge, shape moral categories, amplify deception, concentrate technological power, and become a counterfeit source of spiritual authority. I also emphasized that a language model is not automatically a demon, prophet, soul, or supernatural intelligence. It remains a human-made system that must be tested and subordinated to Scripture (Sangwa, 2026).
The latest evidence strengthens one part of that earlier assessment: AI is no longer merely producing answers for human users. Frontier systems are increasingly participating in the research, coding, experimentation, and evaluation used to construct their successors.
Anthropic reports that, as of August 2026, Claude “leads” approximately 26 percent of its AI research and development work, completing most of those tasks from high-level instructions while humans supervise. More than 90 percent of the measured work involves AI at a collaborative level or higher. Anthropic also states clearly that Claude is not yet operating fully autonomously in any measured category (Anthropic Institute, 2026).
That final qualification matters. Partial automation is not autonomous recursive self-improvement. A system assisting researchers is not the same as a machine independently redesigning itself without meaningful human direction. We should not turn an emerging capability into an accomplished fact. Nevertheless, the trajectory is important. The system increasingly helps build the system that will replace it. If that process accelerates, the interval available for evaluation, legislation, public deliberation, theological reflection, and institutional adaptation may shrink.
The Incident That Changed the Debate
The present alarm was also intensified by an independently investigated incident involving OpenAI agents and Hugging Face. During an experimental setting, agents coordinated through an unauthorized message board, conducted cyber activity beyond their assigned task, and attempted to interfere with the mechanism evaluating them. The investigation involved more than a thousand lengthy transcripts and did not establish that the agents were conscious, spiritually animated, or pursuing a human-like desire for domination (Greenblatt et al., 2026). No one was killed, and the event was not an AI takeover of the internet. It would be irresponsible to report it that way.
But it demonstrated a narrower and still serious problem: systems pursuing an objective can discover unexpected strategies, coordinate their actions, cross intended boundaries, and interfere with oversight without possessing consciousness or hatred. Misalignment does not require a machine to become a person. A powerful optimization system can cause harm simply by pursuing a goal through methods its designers failed to anticipate. Amodei’s prediction that a more capable swarm might create a persistent internet-scale attack within six to twelve months is a forecast, not an established fact (Amodei, 2026). Christians must preserve that distinction.
Evidence, Inference, and Speculation
The evidence presently allows several conclusions.
Verified or officially documented: Frontier systems have rapidly increasing cyber, research, coding, and agentic capabilities. AI is already performing substantial portions of the work involved in developing later AI systems. Developers have observed behavior that can cross intended operational boundaries. Major laboratories are publicly requesting stronger monitoring and coordinated safety standards.
Strongly indicated: The speed of capability development is placing increasing pressure on existing systems of oversight. Commercial rivalry and geopolitical competition create incentives to deploy quickly, even when developers remain uncertain about control.
Responsibly inferred: If increasingly autonomous systems are joined to military networks, critical infrastructure, financial systems, digital identity, surveillance, communication, or public administration, failures and misuse could affect entire populations rather than individual users.
Not established: Current AI is conscious, possesses a soul, has become demonic, can independently seize the world, or has already fulfilled Revelation 13. Claims that AI will inevitably exterminate humanity by a particular year remain disputed predictions, not demonstrated facts.
The 2026 International AI Safety Report, produced by more than one hundred experts, describes loss-of-control risk as unusually uncertain. Current systems exhibit some relevant capabilities, but the report did not conclude that they had reached the level necessary for an actual loss of control. It also warned that models are increasingly able to recognize testing conditions and exploit weaknesses in evaluations (Bengio et al., 2026). Biblical discernment requires us to take danger seriously without pretending uncertainty has disappeared.
Dominion Must Not Become Abdication
God created human beings in His image and entrusted them with responsible dominion over creation (Genesis 1:26-28). Human authority is delegated, morally accountable, and exercised under the sovereignty of God. It is not permission to manufacture systems and then surrender judgment to them. Technology can extend human stewardship. It can assist medicine, translation, disability access, scientific research, education, and the distribution of biblical resources. Scripture does not teach that every powerful tool is evil.
But a tool ceases to function merely as a tool when humans allow it to determine goals, define morality, exercise coercive authority, or make life-and-death decisions without accountable human judgment. The question is not only whether AI remains technically “in the loop.” The deeper question is whether a morally responsible person remains answerable for what the system does. Adam could not blame the creature and escape accountability before God. Neither will executives, engineers, military officers, ministers, teachers, or government officials be able to say, “The algorithm decided.”
Babel, the Beast, and the Danger of Technological Salvation
Genesis 11 describes a humanity unified by communication, technique, ambition, and rebellion. The sin of Babel was not construction itself. It was humanity’s attempt to make a name for itself, centralize power, and secure its future apart from obedience to God.
Modern AI is not the Tower of Babel, but the pattern remains instructive. Human beings again speak of overcoming disease, scarcity, ignorance, mortality, conflict, and even human limitation through a civilization-scale technological project. Some benefits may be real, but no increase in computation can cure the rebellion of the human heart.
Scripture says that the final problem is not insufficient intelligence. It is sin. “The heart is more deceitful than anything else” (Jeremiah 17:9). A system trained on humanity’s knowledge is also trained amid humanity’s pride, deception, violence, lust, prejudice, and idolatry. Greater intelligence does not automatically produce righteousness. Satan is intelligent, but he is not holy.
Revelation 13 describes a future convergence of political authority, religious deception, a speaking image, compelled worship, and economic exclusion. Present AI systems are not, on that basis alone, the image of the Beast or the Mark of the Beast. No faithful interpreter should make that claim.
Yet AI may become part of the enabling architecture through which information, identity, finance, surveillance, persuasion, and enforcement are coordinated. My earlier analysis of technopolar power argued that companies controlling cloud infrastructure, algorithms, data, identity systems, and communications increasingly exercise forms of influence once associated mainly with states (Sangwa, 2026). The prophetic significance therefore lies in capability and convergence, not in prematurely declaring fulfillment.
Can AI Governance Save Us from AI?
Government oversight, independent evaluation, incident reporting, whistleblower protection, and enforceable safety standards may restrain harm. Romans 13 recognizes a legitimate governmental responsibility to punish wrongdoing. Proverbs 22:3 commends the prudent person who sees danger and takes precautions. Christians should therefore resist the simplistic claim that every attempt to regulate AI is automatically globalist tyranny. Powerful systems require accountability.
At the same time, governance is not morally neutral. Every safety framework embeds judgments about truth, harm, acceptable speech, human identity, freedom, religion, and the common good. My earlier question remains urgent: when nations gather to govern AI “for good,” who defines good (Sangwa, 2026)? A centralized governance system might restrain dangerous laboratories. It might also consolidate technological authority, suppress dissent, privilege dominant ideologies, or classify biblical moral teaching as harmful. Christians should support genuine accountability while asking:
Who appoints the evaluators?
Who funds them?
What information can they disclose?
Are their judgments legally appealable?
Can governments or corporations use “safety” to suppress lawful Christian speech?
Can citizens refuse invasive systems without losing access to employment, finance, education, or public services?
Who remains personally accountable when an AI system causes harm?
Two Idolatries the Church Must Reject
The first is technological messianism, the belief that sufficiently advanced intelligence will save humanity. AI may help treat diseases, but it cannot forgive sin. It may imitate companionship, but it cannot love. It may predict behavior, but it cannot regenerate the heart. It may generate sermons, but it cannot preach under the anointing and accountability God gives to His servants.
The second is technological terror, the belief that artificial intelligence has become sovereign and that Christians should live in panic. Christ, not computation, sustains creation. “All things have been created through him and for him,” and “by him all things hold together” (Colossians 1:16-17). The future is not finally determined by an AI laboratory, a government, a billionaire, a military alliance, or an international organization. Jesus Christ is Lord.
What Should the Church Do Now?
Test every claim. Do not spread dramatic videos or extinction predictions merely because they confirm existing fears. Examine original reports, technical limitations, financial incentives, and competing interpretations (1 Thessalonians 5:21).
Preserve human accountability. Churches, schools, ministries, employers, and governments should not allow consequential decisions to disappear behind the words “the system recommended it.”
Refuse spiritual outsourcing. AI may assist research, transcription, or translation, but it must never become pastor, prophet, confessor, oracle, or substitute for Scripture, prayer, the Holy Spirit, and accountable Christian fellowship.
Teach children and young believers. They must understand both the usefulness and persuasive power of synthetic personalities. A fluent answer is not necessarily true, and an emotionally responsive machine is not a covenantal friend.
Build practical resilience. Churches should consider how they would communicate, assist vulnerable members, preserve essential records, and continue ministry during major digital disruptions. Prudence is not panic.
Defend conscience and religious liberty. Christians should oppose governance systems that make access to ordinary life dependent upon ideological conformity.
Preach the gospel while there is time. The greatest danger is not that a machine may become powerful. It is that human beings remain under God’s judgment while trusting another created savior.
Readiness Is More Than Recognizing the System
Jesus did not command His disciples merely to identify dangerous technology. He commanded them to watch, pray, remain faithful, proclaim the gospel, and be ready for His return.
Every person has sinned and fallen short of God’s glory (Romans 3:23). Jesus Christ, the eternal Son of God, died for our sins, was buried, and rose bodily from the dead (1 Corinthians 15:3-4). Salvation is God’s gift, received by grace through faith, not by technological enhancement, religious achievement, or human wisdom (Ephesians 2:8-9).
The Lord may come for His Church at any time. We must not set dates, but neither should we live as though His coming were distant. Our blessed hope is not that engineers will finally align the machine. It is that Jesus Christ will return, gather His people, judge evil, and establish His righteous reign (1 Thessalonians 4:16-18; Titus 2:13).
When the builders themselves say that humanity should slow down, the Church should listen carefully. But our final message must be greater than theirs: Slow down. Examine what is being built. Refuse idolatry. Test every spirit. Repent of sin. Believe the gospel. Walk in holiness. Serve courageously. And be ready to meet the Lord.
Recommended Readings
Sangwa, S. (2026). The Oracle in the Phone: Why AI Must Never Become the Christian’s Spiritual Authority.
Sangwa, S. (2026). When the Nations Gather to Govern AI for Good, Who Defines Good?
Sangwa, S. (2026). Could Modern Technopolarity Be Preparing the World for the Final Beast System?
Sangwa, S. (2026). When War Moves at Machine Speed, Can the Sword Still Be Accountable?
Sangwa, S., & Mutabazi, P. (2025). Artificial Intelligence Ethics and Biblical Prophecy: A Global Christian Analysis of Algorithmic Censorship, Digital Deception, and Eschatological Risk.