25 Aug 2025

AI News Digest

🤖 AI-curated 8 stories

Today's Summary

AEye teaming up with Blue-Band to sprinkle Apollo lidar into traffic infrastructure is a solid move towards smarter cities. It’s like taking lidar out of the car and using it to make intersections smarter — a shift that could really enrich real-time data for city systems and autonomous vehicles. Meanwhile, MIT’s new benchmark for measuring AI’s ‘emotional intelligence’ is a fascinating pivot towards understanding the social impact of AI, which could really change how we evaluate these systems in the future.

On the hardware side, Cearvol’s Wave Lite hearing aid is another step in AI’s journey into consumer healthcare, packing some serious tech to make life a bit easier for those with hearing loss. And in the software world, Microsoft’s Project Ire is quietly making waves by leveraging AI to hunt down malware, potentially reshaping the way security teams handle threats. Together, these stories highlight how AI is weaving into both our cities and our personal tech, quietly reshaping the landscape.

Stories

AEye inks partnership with Blue‑Band to embed Apollo lidar into AI traffic infrastructure

AEye announced a strategic partnership with Blue‑Band to integrate its Apollo long‑range lidar and OPTIS™ perception stack with Blue‑Band’s Integrator‑AI™ traffic orchestration platform. The tie-up aims to deploy combined sensing + AI capabilities for traffic monitoring, incident detection and real‑time intersection controls — a push to move lidar from vehicle perception into city infrastructure. Why it matters: the deal shows lidar vendors shifting toward smart‑city and infrastructure use cases (not just vehicles), and could accelerate deployments that feed richer, real‑time data to municipal systems and autonomous fleets.
Read more → Business Wire

Cearvol launches Wave Lite — an AI‑powered in‑ear hearing aid with 40dB gain

Cearvol announced Wave Lite, a new in‑the‑ear hearing aid that uses on‑device AI for acoustic management (noise reduction, feedback cancellation, own‑voice reduction) and delivers up to 40dB gain. The device includes Bluetooth streaming, four scene modes, fast charging and IPX5 resistance, and is being sold initially via Cearvol’s website. Why it matters: it’s another example of AI being embedded into consumer healthcare devices to improve accessibility and real‑world performance — an increasingly crowded market where software differentiation matters as much as hardware.
Read more → PR Newswire

MIT Researchers Propose a Benchmark to Measure AI 'Emotional Intelligence'

WIRED reports (based on a paper circulated by MIT Media Lab) that researchers have proposed a new benchmark to evaluate the emotional and social impact of large language models. The benchmark focuses on how models influence users — e.g., encouraging healthy social habits, fostering critical thinking and creativity, and avoiding emotional dependence or manipulation. This marks a shift in research evaluation from purely task‑performance metrics toward human-centered measures that capture psychological effects of AI systems. If adopted, such benchmarks could reshape model evaluation and safety practices across industry and academia, and influence vendor model‑posttraining and deployment choices (e.g., personality tuning, safeguards for vulnerable users).
Read more → WIRED

AINL‑Eval 2025: Shared Task and 52k‑sample Dataset to Detect AI‑Generated Scientific Abstracts (Russian)

An arXiv preprint (AINL‑Eval 2025) describes a shared task and public dataset for detecting AI‑generated scientific abstracts in Russian. The dataset contains 52,305 samples (human abstracts across 12 scientific domains plus AI‑generated counterparts from five state‑of‑the‑art LLMs). The task was run in two phases with participating teams and aims to test detectors’ robustness to unseen domains and unseen generator models. By releasing a multilingual, large-scale benchmark and continuous shared‑task platform, the work directly addresses academic‑integrity and model‑detection research gaps and should spur new detection techniques and evaluation standards for multilingual scientific content.
Read more → arXiv

Thoma Bravo to buy Verint in $2B deal — a private‑equity play on AI‑powered customer experience

Private‑equity firm Thoma Bravo announced an agreement to acquire customer‑experience automation vendor Verint for about $2 billion (including debt). The deal highlights PE’s continued appetite for software assets that can be re‑engineered around AI and cloud subscription models. For the AI industry this signals further consolidation in contact‑center/CX software, potential acceleration of Verint’s shift to AI‑native and cloud offerings, and heightened competitive pressure on rivals that supply enterprise AI agents and CX automation.
Read more → Reuters

Pintarnya nets $16.7M Series A to scale AI job‑matching and fintech services for Indonesia’s blue‑collar workforce

Indonesian startup Pintarnya raised a $16.7M Series A to expand an AI‑driven employment marketplace that also offers tailored financial services (secured loans, upskilling, side‑gigs) for informal and blue‑collar workers. The round — led by Square Peg — underscores investor interest in regionally focused AI applications that combine marketplace dynamics with embedded fintech. The funding should let Pintarnya scale its matching algorithms, deepen lender partnerships, and push toward profitability while illustrating how generative/ML tools are diffusing into emerging‑market labor and fintech stacks.
Read more → TechCrunch

Microsoft’s Copilot app for Windows 11 gets a fresh UI and practical modules

Microsoft rolled out a major Copilot for Windows 11 update (rolling since Aug 20, 2025) that redesigns the app home and adds practical modules — Recent Files, Copilot Pages (persistent research projects), an app-launcher with Copilot Vision overlays, and conversation history. The update tightens OS integration (file access, vision-backed guided help, memory prompts) to make Copilot a more useful day-to-day assistant for productivity and coding workflows. Impact: improves developer and knowledge-worker productivity by making AI assistance quicker to access and act on local files and apps, and signals Microsoft’s push to bake Copilot deeper into Windows UX.
Read more → Windows Central

Microsoft’s 'Project Ire' AI agent can reverse‑engineer and detect malware

Microsoft quietly launched Project Ire, an autonomous AI agent (announced in tests published Aug 22, 2025) that reverse-engineers software files to detect and classify malware. By combining LLM reasoning with tools like angr and Ghidra and a built‑in validator, Project Ire produced high precision in tests and even produced evidence strong enough for Windows Defender to auto-block an APT strain. Impact: this is a practical, high-value AI tool for security teams — it can accelerate threat triage, reduce analyst burnout, and change how enterprises scale malware detection and response.
Read more → IT Pro