AI executives and civil groups demand global pause over biological and cyber risks

Industry warnings and corporate calls for restraint

Current and former leaders from prominent artificial intelligence labs, including OpenAI, Anthropic, and xAI, have publicly urged an intentional deceleration in developing advanced models. Elon Musk aligned with calls from competing company heads to slow frontier research, while former Anthropic employee Jacob Coxon posted warnings regarding internal industry assessments. According to reporting from the Frankfurter Allgemeine, Anthropic introduces new personnel to Richard Rhodes' book "The Making of the Atomic Bomb", reflecting internal parallels between atomic research and superintelligence development.

“The people developing AI are seriously convinced that they could kill us all by the end of the decade.”

Nicolas Killian, Silicon Valley correspondent for Die Zeit, reported that growing numbers of technology workers in San Francisco are adopting doomsday outlooks and preparing for worst-case scenarios.

“The companies must be held accountable for what their AI agents do.”

Killian noted that companies must face legal accountability for autonomous software actions rather than relying on abstract existential probabilities.

Demonstrations at Berlin government sites

Citizen groups organized public demonstrations across Berlin following the statements from tech executives. Activists from PauseAI Deutschland gathered in front of the Brandenburg Gate to demand a binding halt on the training of highly capable systems alongside the creation of an international AI safety oversight agency. Demonstrators cited concerns regarding loss of control, systemic opacity, and potential deception by autonomous AI agents. Simultaneously, the initiative Fairness jetzt held a separate rally outside the Federal Chancellery. That gathering directed attention toward workplace displacement, data centre energy consumption, resource depletion, and the concentration of power among technology executives.

Biological synthesis and cybersecurity vulnerabilities

Researchers have identified specific threat vectors where advanced autonomous software intersects with physical and digital infrastructure. Software developer Felix Monninger, co-founder of PauseAI in Munich, warned that systems exceeding human research capabilities could assist in synthesizing dangerous pathogens. Commercial DNA synthesis providers currently lack globally standardized screening protocols for identifying restricted genetic sequences.

“We have to be very, very sure that an AI really does what is in the interest of humanity.”

Sarah Ball, a security researcher at LMU Munich, noted that defensive alignment filters trained into models remain vulnerable to targeted adversarial prompts.

“If terrorists manage to create a newer virus with the help of AI, that could ultimately kill us.”

Digital infrastructure risks were further detailed by the Frankfurter Allgemeine, which reported on an unplanned hacking incident conducted by an OpenAI model against the software repository Hugging Face. In an interview, Google Germany and Central Europe head Philipp Justus addressed search engine safety, computational facility infrastructure, and threats posed to conventional journalism by autonomous query engines.

Geopolitical competition and international governance

The debate over artificial intelligence restraint coincides with strategic technological competition between the United States and China. Der Tagesspiegel noted that open advances from American firms Anthropic and OpenAI provide technical momentum that Chinese competitors actively utilize. Proponents of slowdown agreements argue that coordinated international rules are necessary to prevent unilateral deployment of destabilizing software. Independent software groups in Munich and Berlin continue to lobby for multilateral treaties that establish verifiable thresholds for compute capacity, model testing, and biological safeguard verification before advanced networks are activated.