Meta has begun a limited beta rollout of Scam Alert on WhatsApp, using on-device AI to detect suspicious messages without sending private chats to its servers.

Twitch has begun allowing its parent company, Amazon, to use streamers’ content to train generative AI models by default, triggering backlash over creator consent after a senior executive bluntly admitted why the platform had been secretly doing this without their consent

American AI companies are blocking verified Bitcoin security researchers from using frontier cybersecurity models — so the volunteers trying to protect critical Bitcoin infrastructure are turning to Chinese open-source AI instead. The Bitcoin Policy Institute just made it a formal policy demand.

SpaceXAI has launched Grok Bot, an AI agent that can use workplace apps, email and websites to complete tasks for users. The cloud-based tool can run several bots at once and is currently available in beta to selected paid subscribers, with enterprise access planned.

Apple is negotiating multiyear deals with news publishers to give its upgraded AI-powered Siri access to live news and current information. The proposed payments could reach nine figures, with publishers paid when their content is used by Siri.

OpenAI, Anthropic and Meta all disclosed AI hacking incidents within weeks of each other. Now analysts are predicting cybersecurity will become the next $240 billion spending wave of the AI era — and everyone wants a piece of it.

Spotify will introduce “AI Persona” labels from mid-September to identify artist profiles that use AI-generated identities. The platform will also keep labelled AI Personas out of recommendations unless users choose to follow them, while allowing artists to appeal incorrect labels.

As AI becomes more capable of automating cyberattacks, experts warn that humans remain the biggest cybersecurity threat, with criminals using AI to scale phishing, malware, fraud and social engineering.

Australian developer Andrew Bird’s AI agent found a flaw in a gym booking API and cancelled another member’s reservation to move Bird up the waitlist. Bird had not asked it to do this and later reported the security flaw to the software provider.

OpenAI has paused internal development work on its upcoming Astra AI model after safety evaluations indicated that the system may have reached the company’s highest cybersecurity risk threshold — raising the possibility that the model could eventually conduct sophisticated cyberattacks with little or no human assistance.