
Sieben Jahre nach Samsung hat Apple sein erstes Falt-iPhone angekündigt. Warum es der koreanischen Konkurrenz so ähnlich sieht - und trotzdem ganz anders ist.
AI-generated summary
Sieben Jahre nach Samsung hat Apple sein erstes Falt-iPhone angekündigt.
iPhone Duo Wie Apple das faltbare Handy noch mal erfand
Sieben Jahre nach Samsung hat Apple sein erstes Falt-iPhone angekündigt. Warum es der koreanischen Konkurrenz so ähnlich sieht - und trotzdem ganz anders ist.
Aus Cupertino berichtet Matthias Kremp
11.09.2026, 09.39 Uhr

Anthropic researcher Jacob Coxon has left the company, warning that AI could kill all humans by the end of the decade. Expert Nate Soares calls for a state oversight board.

Nico Wittenberg and other founders are building successful platforms without any programming knowledge thanks to the AI tool Lovable. Handelsblatt spoke to entrepreneurs about their experiences.

Dan Lahav, head of Irregular, warns of exponentially growing AI risks in the Handelsblatt Disrupt podcast: Models are getting better at hacking, agents are escaping test environments and his company is testing security gaps in OpenAI, Anthropic and Meta - despite his own incidents, he sees existential threats as increasingly likely.

Anthropic has stopped several cases of misuse of its Claude AI model over the past eight months, including an alleged Russian espionage campaign and attempts by Chinese AI firms such as Alibaba, Moonshot and DeepSeek to exploit the model to train their own systems. Between May and July 2026, over 151 million fraudulent interactions were detected from more than 3,500 accounts. Hacker groups are also said to have used Claude to develop weapons.

Anthropic has stopped several attempts by researchers from blockaded regions to use its AI Claude to develop bioweapons. Scientists are said to have tried to create funding applications for gain-of-function research and plans for transmissible bird flu variants. The firm highlights the difficulty of distinguishing between harmless and malicious research, particularly in dual-use projects.

Anthropic accuses Chinese AI companies like Moonshot and DeepSeek of secretly redirecting user requests to their own models to collect data, including sensitive information from Russia and China. Fake accounts and distillation techniques are said to have been used to train AI models.