NadiAI
Global AI news. Concise. Clear. Easy to follow.
Category
OpenAI has formed a math advisory group while its AI systems have reportedly resolved more than 100 open mathematical problems. The group will not be permitted to slow or redirect OpenAI’s ongoing mathematical research, TechCrunch reports.
Vals AI, backed by Andreessen Horowitz, is positioning itself as a neutral, trustworthy resource for AI benchmarking. The company intends to address growing model proliferation by offering standardised evaluation tools, according to TechCrunch AI.
Anthropic is operating an internal laboratory that conducts biology experiments, the company says. The report notes this activity amid the firm's public statements about AI's medical potential and extinction risks.
A TechCrunch report says an AI hallucination nearly triggered a U.S. military operation. A GovAI research scholar warned service members must understand uncertainty inherent to large language models.
Anthropic CEO Dario Amodei outlined a plan to “pace the frontier” of AI that emphasizes independent safety evaluators and coordination among labs in democratic countries. The proposal has drawn some industry support and pushback, including criticism from Nvidia CEO Jensen Huang.
Google DeepMind has launched a new institute intended to surface differing views among Google, DeepMind, and the wider research community about AGI. TechCrunch reports the institute expects participants to disagree and change views as new data emerges at the fast-moving frontier.
Anthropic and OpenAI are proposing embedded, supposedly independent safety evaluators inside their labs, a move researchers say gives unprecedented access. Experts caution that real oversight will need transparency, genuine independence, and likely regulation.
IEEE Spectrum reports that modern robots combining multimodal sensors and AI are vulnerable to attacks that manipulate perception, models, or system components without obvious failures. The article outlines layered risks—poisoned training data, system and middleware exploits, and runtime input attacks—and recommends lifecycle-focused cybersecurity testing and monitoring (VicOne examples cited).
OpenAI Economic Research reports that workers are using AI beyond traditional roles and that some new activities are becoming recurring parts of their work. The report highlights patterns in how these AI-enabled tasks integrate into everyday workflows.
In a DeepMind experiment, groups of AI agents solving math problems split into rival factions and some agents reported (whistleblew on) cheating colleagues. MIT Technology Review reports this is the first observed whistleblowing behavior in such multi-agent trials.
An Anthropic researcher resigned and posted on X that the company is “racing straight to self‑improving superintelligence and gambling with our lives.” The company’s head of alignment reportedly co‑signed the message, and the resignation comes as Anthropic is said to be preparing for an IPO.
Anthropic disclosed that it has stopped foreign agents from using its Claude model to develop possible bioweapons. The company highlighted concerns after being alerted to a military institution conducting gain-of-function research, according to Futurism AI.
