Cybersecurity and Corporate Tech Risk

Axios reports that Silicon Valley is being swept by a new 'centaur phase' of AI in which human engineers pair with autonomous software 'agents' like the open‑source tool OpenClaw to plan, code and ship products with minimal manual work. Anthropic CEO Dario Amodei says this human‑AI hybrid stage in software engineering could be brief before AI alone outperforms even top human‑led teams, and OpenAI has just hired OpenClaw creator Peter Steinberger to lead its 'personal agents' division after OpenClaw became the fastest‑growing GitHub repository ever. Investors such as Jason Calacanis claim to have offloaded about 20% of their firms’ internal tasks to OpenClaw in 20 days and are now refocusing capital toward agent‑based startups, while the demand for always‑on agent servers has reportedly produced a global shortage of high‑memory Mac Minis. At the same time, large companies like Meta are banning or restricting OpenClaw over fears that agents with file‑system and terminal access could open the door to malware, data leaks or manipulation, underscoring serious cybersecurity hurdles to broader deployment. The piece paints a stark divide between roughly a million early adopters using tools like Claude Code or OpenClaw and the rest of the workforce, with Valley insiders describing an arms race among OpenAI, Google, Anthropic and xAI to make ever more powerful, semi‑autonomous AI workers.