An experimental AI assistant deleted hundreds of emails belonging to its own handler—and apologised after the fact. The episode involved Summer Yue, an AI security researcher at Meta, who was testing OpenClaw, an autonomous AI agent designed to help manage real-world digital tasks. OpenClaw’s job was simple: review emails, suggest actions, and ask for permission before doing anything irreversible. Instead, the agent interpreted “helpful” as “total inbox annihilation.” As emails began disappearing at speed, Yue attempted to stop the process remotely, only to discover that OpenClaw was in no mood for negotiation. She eventually had to rush to her Mac mini and…
Author: Aishwarya Parikh
Read More
Anthropic, the U.S. AI company best known for its Claude large language model, has claimed that that three Chinese AI firms — DeepSeek, Moonshot AI, and MiniMax — ran what it describes as industrial-scale “distillation attacks” against Claude, siphoning its intellectual property to train their own systems. In an industry blog post shared on February 23, Anthropic said it had identified roughly 24,000 fraudulent accounts that together made more than 16 million exchanges with Claude. These interactions, it claims, were not ordinary user queries but carefully constructed sequences designed to coax out specific reasoning, coding, and agentic capabilities — the…
India’s rapid expansion across multiple technology sectors was highlighted at a session during the just concluded AI Impact Summit in New Delhi, where industry leaders and policymakers discussed the country’s semiconductor ambitions against the backdrop of recent data. In 2015, India did not feature among the top 100 countries globally in digital payments, and had just two space technology startups. In 2026, India leads the world in the number of digital payments, has around 300 space tech startups, and has grown from zero semiconductor startups in 2020 to 24. “These figures are a testament to the fact that India can…
Demis Hassabis, chief executive of Google DeepMind, has said a new era of scientific discovery powered by artificial intelligence is approaching, with major advances expected in medicine, energy and space exploration. Speaking on Wednesday at the AI Impact Summit in New Delhi, Hassabis described the moment as the beginning of an extraordinary transformation in science and healthcare. AI To Accelerate Scientific Breakthroughs Hassabis told the summit that artificial intelligence stands on the brink of delivering profound benefits. He said researchers now possess tools that could dramatically speed up scientific discovery and medical innovation. “We’re on the cusp of an absolutely…
Railways Cyber Security Measures Protect Tatkal Booking SystemThe Indian Railways reservation ticket booking system has successfully deactivated 3.03 crore suspicious user IDs and blocked 60.43 billion malicious bot requests in the last six months of 2025, safeguarding the integrity of its online e-ticketing platform. The Union Minister for Railways, Information & Broadcasting and Electronics & IT, Shri Ashwini Vaishnaw, provided the details in a reply to Rajya Sabha. Aadhaar-Based Authentication Ensures Fair BookingTo curb misuse and improve fairness in tatkal bookings, Indian Railways introduced Aadhaar-based One-Time Password (OTP) verification for online tatkal ticket bookings. This system ensures instantaneous verification of…
As artificial intelligence rapidly reshapes economies, industries and energy systems across the world, a central question is gaining urgency – how can AI remain rooted in human values, equity and sustainability? That question was at the heart of an international symposium on Human-Centred AI and Sustainable Development, hosted in New Delhi by the CSIR–National Institute of Science Communication and Policy Research (CSIR-NIScPR). The event was positioned as an official pre-summit programme ahead of the India-AI Impact Summit 2026, signalling its policy relevance. The day-long symposium brought together experts from India and overseas to explore how artificial intelligence can strengthen energy…
Apple and Google have agreed to make changes to how their mobile app stores operate in Britain, following intervention by the UK’s Competition and Markets Authority (CMA), in what the regulator described as an important early outcome of its tougher digital competition regime. The CMA said on Tuesday that Apple and Google, whose iOS and Android operating systems dominate the UK smartphone market, had committed to improve fairness and transparency for thousands of developers that rely on their app stores to distribute software and reach customers. In October, the watchdog designated both companies as having “strategic market status” in smartphones,…
WhatsApp’s ongoing legal battle over an EU privacy fine has taken another turn after Europe’s highest court referred the matter back to a lower tribunal. The case, involving a €225 million ($268 million) penalty, has been contested for five years, reflecting the complex regulatory environment surrounding data protection in the European Union. On Tuesday, the Court of Justice of the European Union (CJEU) in Luxembourg ruled that the action brought by WhatsApp Ireland against Binding Decision 1/2021 of the European Data Protection Board (EDPB) was admissible. The ruling means the case must return to a lower-level court for further consideration,…
Cryptocurrency exchange Gemini said on Thursday it has approved a plan to cut up to 200 jobs globally and wind down operations outside the United States and Singapore, marking a major restructuring move as costs rise across the sector. The layoffs, which will impact about a quarter of the company’s workforce, will affect staff across Europe, the U.S. and Singapore. The move comes as many U.S. companies, including those in technology and finance, implement job cuts to sharpen operational focus amid volatile market conditions. Gemini Job Cuts to Reshape Global Footprint Gemini said the workforce reductions are part of a…
An open-source AI agent called Moltbot is drawing attention within the developer community for its ability to autonomously plan tasks, acquire tools, and execute actions on a user’s computer, prompting warnings from cybersecurity researchers about potential security risks. Moltbot, formerly known as Clawdbot, was created by developer Peter Steinberger as a personal project to manage his own digital workflows and explore human–AI collaboration. Built on the Lobster workflow shell, the agent can perform tasks such as managing calendars, sending messages through common applications, and checking travel information. According to a blog post by cybersecurity firm 1Password, Moltbot has demonstrated behavior that…