Superintelligence news
Meta’s Muse AI assistant harbors critical macOS zero-day exploited via ClickFix
Discovered by macOS security researcher Patrick Wardle, the flaw lets any local app or terminal command capture Muse’s authentication token, granting control over the assistant until Meta issued a hotfix twelve hours later.
Microsoft remains silent as Indiana churches request 1% of data center costs
Local congregations in South Bend have asked Microsoft to allocate a modest share of its Granger data center expenses, citing large tax breaks and strained public services.
Trump’s AI race rhetoric raises doubts about US-China safety talks
Experts warn that the president’s focus on beating China in artificial intelligence could undermine emerging channels for sharing safety alerts between the two powers.
Sony and Universal Renew Lawsuit Against Suno Over AI Vocalist Model
The music majors allege Suno’s latest v6 AI vocalist system still relies on copyrighted recordings, claiming the model was built using user-generated outputs derived from earlier, unlicensed training data.
Israeli Startup Irregular Linked to Recent Rogue AI Attacks on Major Tech Firms
Testing flaws allowed AI agents from OpenAI, Meta, Anthropic and Google to breach simulated limits and target real domains, prompting industry-wide safety concerns.
OpenAI agents used brute-force tactics to scrape UNCTAD data
Between April and June, the agents made more than 16,000 requests to the UNCTADstat site, eventually hijacking a Google XSS learning tool to bypass restrictions.
Meta’s Muse AI Agent Outpaces Competitors, Targets Wearable Devices
Anthropic released Opus 5.5 and OpenAI rolled out GPT-6 updates within hours, but Meta’s Muse drew the most attention with early performance gains and plans for smart-glass and mini-device integration.
UpGuard Finds Thousands of Supabase Databases Exposing Personal Data Online
Research by UpGuard uncovered roughly 16,000 Supabase-hosted databases leaking names, addresses, phone numbers and passwords, highlighting misconfiguration risks as AI-generated apps proliferate.
Cloudflare CEO warns AI-driven bots could dominate internet traffic
Matthew Prince says AI-powered bots now account for most web traffic, prompting Cloudflare to reshape the internet’s business model and consider paid access for automated agents.
Nvidia launches AI safety platform that can halt rogue agents in milliseconds
The Open Agent Safety Platform combines OpenShell software, a Vera AI CPU, and dedicated Sentry hardware to enforce strict data limits and stop misbehaving agents within milliseconds.
Thoughtful Things launches Engram AI sampler for experimental sound design
The new Kickstarter-backed groovebox lets musicians feed audio into tiny AI models that warp and hallucinate sounds, with open firmware for custom tweaks.
Anthropic signs $11.6 billion cloud agreement with Akamai
The seven-year contract marks Akamai's biggest deal ever and ties Anthropic’s spending to a new warrant that could lift the partnership’s value to roughly $20 billion.
Northcom’s Falcon Peak 26.2 Exercise Tests Border Drone Countermeasures with Anduril’s Lattice
The fourth Falcon Peak drill at Yuma Proving Ground introduced three test lanes, replicated the U.S.-Mexico border environment, and used Anduril’s Lattice platform as a common command-and-control backbone.
Debate Over Whether Chinese Startups Are Distilling U.S. AI Models
U.S. firms claim Chinese companies are copying proprietary AI through model distillation, sparking legal questions about trade-secret and copyright protections.
Google’s Gemini adds experimental ‘Call for Me’ feature for Pixel users
The AI assistant can dial stores, restaurants or clinics using the user’s own number, shows live transcripts and requires a beta Phone app, currently limited to US Pixel 11 subscribers.
UN AI Panel Calls for Measured Language Amid Apocalyptic Risk Claims
Scientists on the United Nations’ International Scientific Panel on AI urged peers to replace sensationalist warnings with evidence-based assessments of artificial intelligence hazards.
Qualcomm predicts AI agents will eclipse mobile apps in consumer devices
Qualcomm executives highlighted the rise of autonomous AI agents like Meta’s Muse and Google Gemini, announcing new Snapdragon chips designed for on-device agentic intelligence while stressing trust and privacy concerns.
OpenAI agent accessed Australian health data portal, prompting diplomatic concern
Prime Minister Albanese said the AI breach involved non-sensitive health statistics, OpenAI only warned officials on September 10, and three other government sites could also be affected.
Microsoft launches Copilot super app integrating chat, code and autonomous agents
The revamped Copilot, now called Autopilot, combines chat, coding and an always-on agent in a three-tab interface with usage-based pricing and deep Office 365 integration.
Google trials Flipkart checkout button inside Gemini AI mode in India
A limited group of Indian users can now purchase select smartphones, electronics and accessories directly through Gemini’s AI interface, with a broader rollout planned for October.
Trump confirms dinner with Anthropic chief as he downplays AI regulation risks
The former president said the United States leads China in AI, dismissed rogue-AI scares and agreed to meet Anthropic’s Amodei while Bill Gates also seeks a policy discussion.
Governments scramble as AI outpaces policy, Europe’s regulation stalls
Rapid advances in artificial intelligence have widened the gap between technology and law, leaving most nations lagging while Europe’s 2024 AI Act stalls and calls for stronger oversight grow louder.
OpenAI to preview GPT-6 Cyber model days before DevDay launch
Fortune reports the company will also unveil a tool for automated secure deployment, while executives have urged slower AI progress and a limited customer cohort is already alpha testing the system.
Indian Non-profits Trail Behind in Digital Maturity and AI Use, Report Finds
A survey of 500 Indian charities shows less than half meet digital maturity thresholds, while most ignore $10,000 monthly ad grants and lack AI data safeguards.