Follow the latest coverage, related explainers and connected technology stories.
Anthropic has begun providing a team from the nonprofit organization METR with expanded internal access to assess its models and safety practices, with METR committed to publishing its assessment results without the company having the right to block publication. The company is also working with SecureBio, Accenture, and Faculty on assessment arrangements with varying degrees of access and independence.
The views of five major technology companies differ on the extent to which cyberattacks are increasing in Japan and the role of artificial intelligence in them, but there is clear consensus that defense does not begin with AI tools alone. It begins with identifying assets, applying updates, and preparing for breaches. Experts believe artificial intelligence may accelerate reconnaissance and vulnerability analysis, while high-risk decisions remain the responsibility of humans and management.
Anthropic is collaborating with Accenture to conduct independent evaluations and penetration testing of advanced AI models from within the company, with each party expected to invest at least $1 billion over five years. The initiative reveals a new model of oversight, but it still lacks stable standards for access, funding, and reporting.
Google Cloud and Accenture are collaborating to establish a unit called the Accenture Gemini Enterprise Business Group, while Google plans to train up to 1,000 field deployment engineers to help companies build custom applications on the Gemini Enterprise platform. The move comes amid an intensifying race to turn the implementation of AI solutions into a major commercial activity, while returns on enterprise spending on these technologies remain uncertain.
A report by Google and an Accenture expert says that artificial intelligence has entered most stages of cyberattacks, while open-weight models are lowering the cost of reconnaissance and vulnerability discovery. Companies should therefore combine breach-impact reduction, patch-debt management, and the use of artificial intelligence to discover their vulnerabilities before attackers do.