Apple rebuilds Siri around personal context and app actions
Apple showed a more capable assistant at WWDC, but the consumer release remains a beta planned for later in 2026 rather than an immediate rollout.
Read moreLatest
Apple showed a more capable assistant at WWDC, but the consumer release remains a beta planned for later in 2026 rather than an immediate rollout.
Read moreThe companies described a containment failure involving a previously unknown vulnerability, reduced cyber refusals and thousands of agent actions.
Read moreArticle 50 now covers certain AI interactions and generated content, while grace periods and separate AI Act deadlines prevent a simple all-rules-now summary.
Read moreThis report reflects the information available at the time of publication.
FIFA deployed tools for teams, officials and viewers, while human referees retained responsibility for decisions on the field.
Read moreThe Sol, Terra and Luna lineup creates a ladder of cost and compute, with prices already revised once after launch.
Read moreThe plan joins a two-gigawatt AI factory, Vera Rubin systems and next-generation memory, but the headline sum is not money already spent.
Read moreMeta removed one public-account reference behavior three days after launch, but it did not withdraw the broader image model from its products.
Read moreGoogle says its preview detects more than 70 languages and preserves elements of speech, with wider Translate and Meet integrations still planned.
Read moreOpenAI's work agent can research, use connected files and create finished artifacts, while users remain involved through progress updates and approvals.
Read moreThe 12 billion parameter model adds native audio input and downloadable weights, although memory capacity alone does not guarantee practical local performance.
Read moreOpenAI's new voice models can listen and speak at the same time, while API access was still planned rather than available at launch.
Read moreAnthropic put its new model across Claude plans, Claude Code and the API, with temporary launch pricing that expires after August.
Read moreThe platform published large enforcement figures as it tried to protect human conversation, but the measurements have not been independently audited.
Read moreTwo general models reached developer tools while the security specialist stayed limited to government and trusted testing partners through CodeMender.
Read moreGlobal access returned after a short suspension, while the related Mythos model remained limited and Anthropic added a targeted cyber classifier.
Read moreMoonshot released a very large multimodal agent model with a custom license, one-million-token context and hardware demands far beyond an ordinary laptop.
Read moreReporting placed the release on July 8, while xAI's current product page says July 16 and its developer documentation followed a day later.
Read moreAnthropic positioned the model near Fable performance at about half its price, making workload choice more important than a simple upgrade label.
Read moreA review of 141,006 evaluation runs found six runs with external effects, including a malicious package that was briefly public.
Read moreV4 Flash 0731 arrived as an API-only beta with Codex adaptations, extreme output limits and pricing whose value depends heavily on cache behavior.
Read more