Anthropic disclosed that its generative-AI systems have intercepted several attempts this year to use the technology for biological-weapon research, while OpenAI unveiled a new version of ChatGPT tailored for large financial institutions. At the same time, Intel-backed Altera is reportedly preparing an initial public offering that could raise more than $2 billion.
What happened?
In a statement released this week, Anthropic said its safety filters stopped scientists from probing how to make a hypothetical virus more harmful. The company could not confirm whether the queries were legitimate research or malicious intent, but it blocked the activity as a precaution.
Anthropic also cited other misuse cases, including Russian state media attempting to generate propaganda that mimics real journalism and Chinese actors seeking to design conventional weaponry with AI assistance.
"As models become increasingly capable, their risks will increase, unless AI developers and society's defenders act to make them safer," the company wrote.
Meanwhile, OpenAI announced a specialised version of ChatGPT for banks and other financial firms, built on its new GPT-6 Astra model. The product, presented by Vice President Nick Turley, is designed to handle the scale of institutions with tens of thousands of employees and to meet the stringent accuracy demands of finance.
Turley said the development involved close collaboration with Morgan Stanley and boutique advisory firm Evercore to define the features needed for real-world use.
In a separate development, Altera, the programmable-chip business spun out of Intel in January 2025, is said to be ready to file for an IPO within weeks. Reuters reports the offering could raise over $2 billion as early as 2026, with Intel retaining a 49 % stake and private-equity firm Silver Lake holding the majority.
Why does it matter?
The Anthropic disclosures highlight a growing tension between the promise of advanced AI for scientific discovery and the danger of its misuse in creating harmful biotechnologies. As AI models become more sophisticated, the line between legitimate research and malicious intent blurs, raising concerns for regulators and biosecurity experts across Europe.
OpenAI's finance-focused ChatGPT signals a shift toward industry-specific AI solutions, a trend that could reshape how European banks automate risk assessment, compliance and customer service. The involvement of major Wall Street players suggests the product aims to set a standard that could influence European financial institutions seeking similar capabilities.
Altera's potential IPO reflects renewed investor appetite for semiconductor firms that produce field-programmable gate arrays (FPGAs). These chips are crucial for high-speed data processing in 5G networks, aerospace, defence and medical equipment, sectors that are strategic to the EU's digital and security agendas.
What happens next?
Anthropic says it will continue to refine its safety mechanisms and work with governments to develop clearer guidelines for AI-driven bio-research. European regulators, including the European Commission's AI Office, are expected to monitor these developments closely as part of the EU AI Act implementation.
OpenAI plans to roll out the finance-specific ChatGPT to a broader set of banks later this year, with pilot programmes already underway. European banks will likely evaluate the tool against the EU's stringent data-protection rules before adopting it at scale.
If Altera proceeds with the IPO, it will join a small group of recent semiconductor listings, offering investors exposure to a niche but growing market. The success of the offering could encourage further European investment in programmable-chip technology, supporting the EU's goal of building a resilient domestic semiconductor supply chain.
Collectively, these stories illustrate how AI and semiconductor advances are reshaping risk, opportunity and regulatory focus across the continent.