Switzerland signs the Council of Europe’s Framework Convention on Artificial Intelligence and tasks the FDJP, DETEC, and FDFA with drafting a bill to implement transparency, data protection, non-discrimination, and oversight provisions by end of 2026. Until parliamentary ratification and potential referendum, AI remains governed by existing constitutional, data protection, civil, and criminal liability frameworks to foster innovation, protect fundamental rights, and enhance public trust.

Key points

• Switzerland signs the Council of Europe’s AI Convention, pending parliamentary ratification and possible referendum.
• Federal Council tasks FDJP, DETEC, and FDFA with drafting a bill by end of 2026 covering transparency, data protection, non-discrimination, and oversight.
• Until ratification, AI remains governed by the Swiss Constitution, Data Protection Act, and existing civil and criminal liability statutes.

Why it matters: This move establishes a binding, human-rights-based AI regulatory framework in Switzerland, balancing innovation with fundamental rights and setting a global policy precedent.

F-Prime Capital analyzes worldwide robotics investment trends, revealing an $18.5 billion funding rebound in 2024 and detailing traditional and alternative financing tools, regulatory impacts, and strategic partnerships for early-stage companies.

Key points

• 2024 global robotics investment rebounds to $18.5 billion, driven by 50+ mega-rounds over $50 million.
• Early-stage firms face high R&D and material costs, spurring interest in SBIR/STTR grants, venture debt, and crowdfunding.
• Regulatory factors like CFIUS reviews and DEI executive orders critically affect fundraising timelines and compliance.

Why it matters: Mapping evolving robotics funding channels reveals how startups can secure capital efficiently, driving innovation and maintaining competitive leadership in AI and automation.

In a key decision, the Federal Circuit holds that patents claiming generic machine learning methods applied to existing problems lack eligibility under Section 101, reinforcing that AI inventions must involve genuine technical improvements to secure patent protection.

Key points

• Federal Circuit affirms Section 101 dismissal of ML patent claims applying generic algorithms to scheduling and mapping use cases.
• Court finds no improvement to core machine learning models or training techniques, deeming them “conventional” and ineligible.
• Decision underscores that automating known human tasks with standard ML methods without technical innovation fails patent eligibility.

Why it matters: This ruling establishes a stricter patentability threshold for AI inventions, emphasizing substantive technical contributions over mere applications of off-the-shelf machine learning techniques.

A Federal Circuit panel concludes that patents merely applying generic machine learning to new datasets lack eligibility under the Alice framework, requiring a transformative inventive aspect beyond routine computing.

Key points

• Federal Circuit holds Recentive’s ML Training and Network Map patents ineligible under Alice Steps 1 and 2.
• Claims reference generic ML models trained on historical event, venue, and weather datasets without technical detail.
• Patents lack inventive concept as they recite conventional computing components and broad machine learning limitations.
• Court emphasizes that efficiency gains alone cannot convert an abstract idea into patent-eligible subject matter.
• Affirms district court’s denial of amendment as any changes would remain technologically conventional.

Think about automatic TV lineup tools: you might expect an AI patent for tuning schedules. But the Federal Circuit found that merely using off-the-shelf machine learning to generate network maps or schedule events—tasks once done by hand—still qualifies as an abstract idea under §101. For example, Recentive’s patents on dynamically training models for NFL game scheduling were deemed generic. Courts said you have to show improvements to the algorithm itself to secure patents.

Key points

• Generic applications of off-the-shelf machine learning in new environments are abstract ideas and patent-ineligible under §101
• Recentive’s broadcast scheduling and network map patents lacked specific technical improvements to their ML algorithms
• Successful AI patents must show concrete algorithmic enhancements beyond standard ML use