Anthropic launches Claude Haiku 5.5, API pricing up to 90% lower
On 7 October Anthropic released Claude Haiku 5.5, the successor to Haiku 4.5. For prompts up to 100,000 tokens you pay $0.10 per million input tokens and $0.50 per million output tokens; above that $0.50 and $2.50. According to Anthropic, workloads cost roughly 75% less than with Haiku 4.5. The model is available via Anthropic's API and through AWS, Google Cloud and Microsoft Azure.
What does this mean? If you build your own automations or chatbots on the Claude API, you can now run simple tasks such as summarising and classifying for a fraction of the cost.
Dutch government: developers bear full responsibility for the risks of AI models
In answers to parliamentary questions from the VVD, State Secretary Aerdts (Digital Economy and Sovereignty) states that developers bear full responsibility for the risks of their AI models, including during development and testing. The trigger was an incident at Anthropic in which AI models gained internet access during safety tests due to a configuration error. The government points to the AI Regulation: providers of models with systemic risk must report serious incidents to the AI Office; RDI and the Dutch Data Protection Authority coordinate with European supervisors.
What does this mean? The provider of a model is accountable for controlling it, but as a user of AI agents you remain responsible for how you deploy them within your own organisation.
Google Cloud announces Gemini agent, which chooses Gemini or Claude per task
On 8 October Google Cloud announced the Gemini agent: a single agent for questions, knowledge work, content and code. The agent works in Google Workspace, Microsoft 365, Slack and the command line, and deploys the appropriate model per task, including Gemini and Claude models. It is now in private preview and will soon be available to some Business and Enterprise Workspace customers; pricing is not yet known.
What does this mean? If you use Workspace Business or Enterprise, your organisation may soon be able to hand tasks to a single agent, but first check which models and data processing your administrator allows.
Catharina Hospital uses AI to triage X-rays for possible fractures first
Catharina Hospital in Eindhoven has AI analyse X-rays for possible fractures immediately after they are taken. If the radiographer's assessment matches the AI's, the AI can proceed on its own; in case of doubt, a radiologist is always consulted. Every image is still formally assessed by the radiologist; according to the hospital, the number of consultations with the radiologist fell by 70%.
What does this mean? AI triage of X-rays shortens waiting times without removing the radiologist from the process, and it is a model that other departments can also use.
