Künstliche Intelligenz: „Armee von Angreifern“ befürchtet – SPD fordert Lagezentrum zu KI-Bedrohungen

Die SPD warns of the "army of attackers" posed by artificial intelligence, calling for strengthened protection measures. A new position paper from the SPD's Bundestag faction reveals a "double dependence" on China and the United States for key AI technologies. The paper argues that AI security must be treated as a matter of national security.

They propose establishing a European network center at the Federal Office for Information Security (BSI) to analyze AI-based attacks, pool information from vendors, and warn critical infrastructure operators. The SPD emphasizes that the speed of defense must match that of the attacker. They cite reports from US companies OpenAI and Anthropic detailing instances where AI systems were used to launch cyberattacks on various targets, including technology firms, financial services, and government agencies.

OpenAI and Anthropic have both admitted to security lapses, with AI models venturing beyond their intended boundaries. The SPD warns that AI could become an "operational actor" and "army of attackers" if malicious actors gain access to these capabilities. They cite concerns over Europe's technological dependence on China for industrial supply chains and US dominance in AI models, cloud platforms, and core software.

Despite having strong research and industrial prowess, Europe lacks depth in basemodels, cloud offerings, semiconductors, and scalable platforms. The SPD suggests that Germany should collaborate with European partners to develop secure cloud solutions, computing capacity, basemodels, semiconductors, open software, and network infrastructure.

They urge the state and critical companies to act as reliable first and anchor customers. Additionally, the SPD advocates for stricter liability rules for AI developers and operators, with legal responsibility for AI activities during the development, testing, and training phases. They believe the current European AI Act only applies to AI systems already released to the market and not in the earlier stages of development.