noticias radar ia precios ia agentes ia ai act

Yesterday's AI Radar called for treating every agent like an employee with keys; today, the three headlines paint the same board from the business side: AI is becoming cheaper at a breakneck pace, agents are starting to do real work—squaring figures, drafting reports—and this Sunday, the European regulation moves from having obligations to being able to fine. OpenAI cuts the price of its cheapest model by 80%, Workiva launches agents that reconcile numbers and draft sustainability reports with an audit trail, and on August 2, the European Commission debuts sanctioning powers of up to 15 million euros. The takeaway of the day: the cost of automating is going down, and the cost of doing it wrong is going up.

Automating with AI costs one-fifth today compared to three weeks ago

On July 30, OpenAI cut the price of its cheapest GPT-5.6 models: Luna drops by 80% (from $1 to $0.20 per million input tokens and from $6 to $1.20 for output) and Terra by 20% (to $2 and $12); the flagship Sol remains at $5 and $30. This is the second reduction in three weeks since the July 9 launch, and the company attributes it to increasingly cost-sensitive customers and competition. For your company: tasks that weren't cost-effective to automate—mass email classification, summarizing tickets, translating product sheets, extracting data from documents—now cost a fraction. So, dust off that automation you shelved and set the cheapest model as the default for routine volume, reserving the powerful one only for what truly needs it. Source

Agents are already squaring figures and drafting reports, not just responding

Workiva presented three agents for corporate closing and reporting on July 30: one checks that every figure in a financial report matches its source and generates explanations for deviations, another compares your wording with that of peer companies, and a third drafts and scores sustainability reports against ESRS and ISSB standards—all with a traceable audit trail and a layer that learns from your own historical documents. For your company: even if you don't use Workiva, the signal is what matters: the next wave of agents is targeting tedious and verifiable back-office work—reconciling numbers, cross-referencing documents—and leaving a record of what they did. When you test an agent, demand that it notes what it touched and against which source, because that trail is what turns a demo into something you can defend before an auditor or a client. Source

On Sunday, the AI Act stops warning and starts fining

On August 2—this Sunday—the European Commission's supervisory and sanctioning powers over general-purpose AI providers are activated: fines of up to 3% of global turnover or 15 million euros (whichever is higher), plus the authority to demand documentation, evaluate models, and order corrections or withdrawals. The one-year grace period that began in August 2025 ends: the regulation moves from existing to being enforceable. For your company: the fines target large providers, not your SMB directly, but there are two practical consequences: those providers will pass documentation and compliance requirements on to you (expect new clauses and forms), and the transparency obligations that do apply to you—labeling generated content, warning that your chatbot is AI—now live under a sanctioning regime. So this weekend is the last quiet moment to finish those quick fixes. Source

What to watch for tomorrow?

The deadline expires on Sunday: attention turns to the first practical guidelines from the European Commission and the AEPD on how to label content daily, and whether other labs follow OpenAI in cutting prices. And the question that brings you back to the beginning: if automating costs one-fifth today compared to three weeks ago, what process that you dismissed as too expensive is now back on the table?

Watch the 1-minute video