citiesabc, first_page
Can AI Really Cause a Billion Deaths? Technology Expert David Linthicum Challenges Bill Gates’ Warning
26 Oct 2026
AI, cloud and cybersecurity expert David Linthicum argues that the debate over catastrophic artificial intelligence risks should focus on practical safeguards, including security architecture, access controls, monitoring and governance.
Could artificial intelligence become powerful enough to cause a billion deaths, or are extreme predictions distracting attention from the security risks organisations already face?
Global technology expert David Linthicum is challenging a warning attributed to Microsoft co-founder Bill Gates, arguing that the AI safety debate needs greater technical scrutiny and a stronger focus on practical risk management.
Linthicum, an AI, cloud computing and cybersecurity specialist, questioned the assumptions behind catastrophic AI scenarios and called for greater attention to engineering, security and governance.
Bill Gates is making a huge leap from what AI is today to a scenario in which it causes a billion deaths,
Linthicum said.
AI is extraordinarily powerful, but it remains a technology that can be engineered, secured, monitored and governed.
His comments offer a different perspective on the growing debate surrounding advanced AI systems, their potential misuse and the responsibilities of businesses and governments developing safeguards.
Catastrophic AI Predictions Require Technical Scrutiny
The discussion follows a warning attributed to Gates that artificial intelligence could become powerful enough to drive events resulting in a billion deaths.
Linthicum argues that catastrophic scenarios involving AI-assisted biological attacks or other large-scale threats depend on several conditions and failures occurring together. Rather than dismissing such possibilities entirely, he believes they should be examined through realistic technical pathways.
The question shouldn’t be whether we can imagine a catastrophic scenario,” Linthicum said. “We can imagine almost anything. The question is: What is the actual pathway from today’s systems to that outcome, what controls exist along the way, and where can we intervene?
His argument centres on distinguishing between hypothetical consequences and demonstrable vulnerabilities. Understanding the systems, permissions and processes that would enable a particular threat, he suggests, is essential to developing effective protections.
This perspective does not establish that catastrophic AI risks are impossible. Instead, it calls for evaluating their plausibility and identifying opportunities for intervention before serious failures occur.
Why AI Security Needs More Than a Kill Switch
One of Linthicum's central arguments is that discussions about whether advanced AI systems require an emergency shutdown mechanism can overshadow more immediate security priorities.
He believes organisations should prioritise systems that are secure, auditable and subject to appropriate access restrictions from the beginning.
The kill-switch conversation is largely a distraction,” Linthicum said. “What matters is building systems that are secure and auditable from the beginning.
For businesses integrating AI into their operations, this includes understanding what individual systems are permitted to access, how their actions are recorded and what restrictions apply to automated decisions.
Monitoring and auditability become particularly important as AI applications move beyond generating information and begin interacting directly with enterprise software, databases and external services.
In these environments, safeguards need to be integrated into the technical architecture rather than treated as an additional feature after deployment.
Cyberattacks, Fraud and Autonomous Agents Present Immediate Risks
While debates about existential threats continue, Linthicum identifies several security challenges that organisations are already confronting.
These include AI-enabled cyberattacks, financial fraud, misinformation, privacy breaches and vulnerabilities involving autonomous AI agents.
Cyberattacks, fraud, misinformation, privacy failures, insecure AI agents and the misuse of AI are real issues that organizations need to address right now,
Linthicum said.
As artificial intelligence becomes more capable, malicious actors may use these technologies to automate parts of existing attacks or increase the sophistication of deceptive content.
At the same time, organisations deploying AI agents must consider whether those systems can access sensitive information, execute unauthorised actions or operate beyond their intended permissions.
Linthicum argues that these challenges can be approached through established security disciplines, supported by stronger governance and monitoring practices adapted to AI systems.
His position places responsibility on organisations to understand and manage the operational consequences of increasingly autonomous technology.
Can AI Be Governed Without Restricting Innovation?
The debate also raises broader questions about the role of governments, regulators and technology companies in establishing safeguards for advanced AI.
Linthicum advocates security architecture, access controls, auditability and governance as practical foundations for managing risks.
These measures are particularly relevant as organisations adopt AI across business operations, customer services, cybersecurity and decision-making processes.
Rather than framing AI development exclusively as a choice between unrestricted innovation and catastrophic danger, his approach emphasises creating systems that remain observable and constrained.
The distinction is important for businesses seeking to benefit from AI while managing legal, security and operational responsibilities.
His analysis also suggests that conversations about AI-designed biological weapons and other severe threats should remain grounded in what existing systems can actually do, the barriers to misuse and the safeguards available to reduce risk.
A More Practical Approach to AI Safety
Linthicum's comments highlight a central tension in the AI safety debate: how much attention should be devoted to possible future catastrophes compared with the vulnerabilities affecting systems today?
His approach calls for assessing risks through technical evidence, identifying the conditions that could allow harmful outcomes and implementing controls capable of interrupting those pathways.
For organisations adopting AI, this means treating security, governance and accountability as essential parts of technology deployment.
Although the potential consequences of increasingly powerful AI systems remain an important subject of discussion, Linthicum maintains that measurable risks and practical engineering solutions deserve greater attention.
The broader challenge is to ensure that artificial intelligence continues advancing within systems that can be monitored, secured and governed, while allowing businesses and society to benefit from the technology's capabilities.
About David Linthicum
David Linthicum is a technology specialist, author and commentator focused on artificial intelligence, cloud computing, cybersecurity and enterprise architecture.
Until early 2024, he served as Managing Director and Chief Cloud Strategy Officer at Deloitte Consulting LLP.
With more than 30 years of enterprise technology experience, he has advised major corporations, technology companies and government agencies on digital transformation and technology strategy.
Linthicum has authored more than 17 books, thousands of articles and over 70 courses for LinkedIn Learning.
