Big Four consulting, Nobel-winning protein prediction, and open-source security — today's AI news is less about what models can do and more about who's responsible when they get it wrong.
GPTZero identified AI hallucinations in four PwC Middle East consulting documents, extending the detection run that previously led EY and KPMG to retract similar work. PwC has not responded publicly.
Google DeepMind has relocated the majority of AlphaFold's original paper authors to other projects, with roughly a quarter of core full-time contributors having left the company.
OpenAI published an early-access Codex Security CLI — open-source tools for scanning code repositories, verifying patches, and adding security checks to CI/CD pipelines.
Anthropic published research showing Claude can independently identify exploitable cryptographic weaknesses in real-world implementations — including subtle constant-time errors that human reviewers miss — without being given specific targets.
CXMT's listing documents disclosed Alibaba owned approximately 5% of the Chinese memory chipmaker before its Shanghai debut; at a ~$464 billion closing market cap, that stake has returned roughly 20x.
ZuriQ, an ETH Zurich spinout, closed a $25.5 million seed round led by Quantonation for its 2D trapped-ion quantum computing architecture — a design intended to overcome the scaling limits of linear ion-trap systems.