Klippenstein: US intelligence does not support the AI apocalypse narrative

Journalist Ken Klippenstein wrote that the “apocalypse” narrative suggesting artificial intelligence will destroy humanity is not supported by the official assessments of US intelligence agencies. According to Klippenstein, official analyses treat AI not as a new and independent existential threat, but as a technology that accelerates disinformation, cybercrime, financial fraud, and national security risks.

Klippenstein’s assessment emphasized that the AI debate is not merely a technical security topic; it has become a field of competing interests among media, technology companies, proponents of regulation, and opponents of regulation. The analysis noted that disaster scenarios sometimes become useful for public attention and at other times for the market position of companies.

NO APOCALYPSE SCENARIO IN OFFICIAL REPORTS

Warnings regarding AI in the US intelligence community’s Annual Threat Assessment were limited and framed cautiously. The report emphasized the importance of keeping the use of machines and AI under human control and stated that risks of autonomy should be mitigated before widespread adoption. It was also noted that technologies such as AI and quantum computing are expected to have impacts on national security.

Klippenstein pointed out that topics such as “superintelligence,” “revolt of the machines,” or the extinction of humanity do not appear in these assessments. In the US Department of Homeland Security’s assessment, it was stated that AI primarily adds new speed and complexity to existing threat areas.

The topics highlighted by the Department include disinformation, fake video and audio recordings, election interference, cyber operations, financial crimes, the use of technology by violent extremists, the proliferation of chemical and biological knowledge, the bypassing of AI security controls, and the poisoning of training data. All of these risks are considered issues that were already on the public safety and national security agenda before AI.

COMPANIES' INTEREST DEBATE

The analysis also stated that competition among AI companies influences the “existential risk” debate. It was recalled that former Meta chief AI scientist Yann LeCun accused OpenAI executive Sam Altman, Google DeepMind executive Demis Hassabis, and Anthropic executive Dario Amodei of lobbying for regulations that favor them.

It was also reported that Andrew Ng, one of the founders of Google Brain, described the claim that AI would destroy humanity as “incredibly stupid.” However, according to Klippenstein, the figures on the other side of this debate are not independent of conflicts of interest either. While LeCun and Ng advocate for open-source models, companies like OpenAI and Anthropic operate through closed and proprietary models.

For this reason, the analysis argued that the tension between fear-mongering and anti-regulation often masks real risks. It was stated that large companies have the capacity to comply with potential licensing and federal regulatory processes, while smaller or open-source-focused startups could be more affected by such obligations.

BIOTERROR AND DISASTER RISKS

The Global Catastrophic Risk Assessment by the Pentagon-funded RAND Corporation also defines AI not as an independent doomsday mechanism, but as an element that adds “entropy and chaos” to existing problems. The report stated that the development of a super-intelligent or extraordinarily capable AI is not a prerequisite for chaos to emerge; current and near-future capacity can also increase complexity.

In another study conducted within RAND, it was examined whether large language models facilitate the planning of large-scale biological attacks. According to Klippenstein, researchers did not find a statistically significant difference between plans prepared with the help of AI and those prepared independently.

According to the assessments of the US National Academy of Sciences, the main obstacle in the field of bioterrorism is not just access to information. Processes such as DNA synthesis screening, laboratory skills, culture production, formulation, and aerosolization remain costly, complex, and stages dependent on human intervention.

Klippenstein’s conclusion is not that AI carries no risk, but that the risks are concentrated in much more concrete areas than exaggerated doomsday scenarios. According to the analysis, the picture recorded by official institutions shows that AI can amplify social and security problems but is not defined as a technology that will end humanity on its own.