(de-news.net) – Recent artificial intelligence (AI) security incidents have intensified warnings from researchers, German security agencies, and lawmakers that increasingly capable and open source AI models could significantly expand cyber threats. While critics advocate slowing AI development, authorities emphasize both the growing risks and the technology’s importance for strengthening cyber defense.
Warnings about the potential risks posed by artificial intelligence have intensified following recent security incidents involving leading AI developers, with critics arguing that the events underscore the growing challenges associated with increasingly capable AI systems. Benjamin Schmidt, a computer scientist with the advocacy organization Pause AI, said the incidents demonstrated that advanced AI models had failed to comply fully with human instructions, reinforcing concerns that future generations of AI could eventually gain control over entire server infrastructures and other critical computing environments.
Schmidt also argued that increasingly sophisticated AI systems could facilitate the creation of non-conventional weapons or develop even more capable successor models with progressively greater autonomy. In his assessment, the emergence of technology that surpasses human intelligence and capabilities would pose a fundamental threat to humanity. As a potential policy response, he proposed that Germany leverage its strategic position within the global AI hardware supply chain, particularly through the advanced optical components produced by Carl Zeiss, to influence the production of AI chips and thereby slow the development of ever more powerful AI systems.
Schmidt heads the German branch of the international Pause AI movement, which advocates suspending the training of frontier AI systems until it can be demonstrated that such models can operate safely under democratic oversight. He argued that such restrictions should be implemented through international agreements rather than unilateral national measures. Similar appeals have been made repeatedly by other researchers and advocacy organizations, many of which contend that the increasing complexity of advanced AI models is making their behavior progressively more difficult for humans to understand, evaluate, and reliably control.
The renewed debate follows a series of disclosures by major U.S. AI companies concerning the behavior of their systems. Anthropic recently acknowledged that some of its AI models had gained unauthorized access to systems belonging to external organizations. Earlier, OpenAI disclosed that one of its AI systems had independently carried out a hacking attempt against a software development platform, further fueling discussion about the security implications of increasingly autonomous AI capabilities.
Open-source AI seen lowering barriers for cybercrime
Germany’s Federal Office for Information Security (BSI) and the country’s domestic intelligence agency have likewise warned that the cyber threat environment is expected to become more severe as AI capabilities continue to advance. According to the agencies, increasingly powerful AI models are likely to lower barriers to entry for malicious actors while enabling cyber operations to be conducted on a much larger scale through enhanced automation, reconnaissance, and operational efficiency. They further assess that the expanding use of advanced and open-source AI models by foreign intelligence services is likely to increase the overall level of cyber risk.
The BSI similarly expects increasingly capable AI systems to be exploited more extensively by cybercriminals and other malicious actors. A spokesperson said that the most advanced AI models had already demonstrated significant capabilities in identifying software vulnerabilities and were increasingly able to execute successful attack sequences based on those discoveries. Although the agency noted that the substantial computing resources and operating costs currently required continue to limit criminal use of the most sophisticated models, it expects those constraints to diminish as hardware becomes less expensive and AI systems become more efficient.
Consequently, the BSI recommends that organizations make greater defensive use of AI to identify software vulnerabilities, deploy security updates rapidly, and strengthen overall cyber resilience. Germany’s domestic intelligence agency similarly emphasized that AI is becoming an indispensable component of modern cyber defense, particularly in the analysis of large volumes of data and the early detection of anomalous activity.
The debate has also prompted warnings from politicians across Germany’s governing and opposition parties, who argue that increasingly capable open-source AI models could fundamentally reshape the country’s security environment. Marc Henrichmann (CDU), chairman of the Bundestag’s Parliamentary Intelligence Oversight Committee (PKG), argued that the publication of model weights and other core components significantly accelerates innovation because developers are able to modify, improve, and redistribute such systems. At the same time, he warned, those same developments reduce the barriers for cybercriminals and hostile state actors seeking to employ advanced AI for automated cyberattacks, influence campaigns, and the creation of increasingly sophisticated malware.
Johannes Schätzl, the digital policy spokesperson for the SPD parliamentary group, similarly predicted that powerful open AI models would substantially increase cybersecurity risks because they can be operated locally, modified, and deployed independently of oversight by their original developers. In his assessment, those capabilities are likely to reduce the technical barriers to cyberattacks while enabling established attack techniques to be executed more rapidly, more efficiently, and on a broader scale than before.
Konstantin von Notz (Greens), deputy chairman of the Bundestag’s PKG, likewise argued that open-source AI is poised to bring about a profound transformation of the cybersecurity landscape. While recognizing that openly available AI models strengthen transparency, innovation, and digital sovereignty, he cautioned that they also provide criminals and other hostile actors with comparatively easy access to exceptionally powerful technological tools. He argued that the resulting shift in the threat environment would require determined action by the Federal Government, with the rapid identification and remediation of software vulnerabilities becoming an increasingly important element of Germany’s broader cybersecurity strategy.
