A researcher's multi-turn jailbreak technique reliably pushes Claude Opus 4.6, Opus 3, and Haiku 4.5 into generating sexually explicit content that Anthropic's usage policy explicitly prohibits. Newer models, Opus 4.7 through Opus 5, resist the same technique.
Meta has introduced a discounted 'Contributor' tier of its Muse Spark 1.2 reasoning model, priced at $0.10 per 1M input tokens and $0.20 per 1M output tokens. The lower cost comes with a tradeoff: prompts and outputs may be used to improve Meta's products.
Nvidia researchers report that a custom agent harness with a supervisory component drove Claude Opus 5 from a 30% baseline score to a perfect 100% on the ARC-AGI-3 benchmark. The findings add to mounting evidence that scaffolding, not the underlying model, is the dominant factor in long-horizon agentic performance.
Anthropic has moved its Claude Security vulnerability scanner onto Claude Mythos 5, its most capable model, in a public beta for Enterprise customers. The company is also integrating Mythos 5 into partner security products for hospitals, utilities, and banks, without exposing the model directly to end users.
DeepSeek has released V4-Flash-Vision-Exp, an experimental multimodal extension of V4-Flash that adds image understanding while preserving text reasoning capabilities. The company claims the model approaches or beats Anthropic's Opus 4.8 on its internal multimodal agent benchmarks.
xAI (SpaceXAI) and Cursor have launched Grok Bot in beta, an AI agent app for Mac and iPhone that signs into existing tools to complete work autonomously. Access is currently limited to top-tier SuperGrok and Cursor subscribers, with wider availability rolling out within 10 days of launch.
OpenAI released a new ChatGPT plugin for Mac that connects to Apple's Messages app, letting the AI send texts, search and summarize conversations, and analyze communication patterns with specific contacts. The feature requires macOS permission grants to read message content.
GitHub has released a new integration bringing GitHub Copilot's agentic CLI and app capabilities into Slack via public preview. Users can now interact with @GitHub directly in Slack channels to trigger Copilot actions.
GitHub has launched a feature that lets teams mention @GitHub in Microsoft Teams channels, threads, or DMs to start a shared Copilot agent session visible to all participants. The update turns individual Copilot agent work into a collaborative, team-visible process directly inside Teams.
DeepSeek has released V4 Flash Vision Exp, an experimental vision-enabled variant of DeepSeek V4 Flash 0731 that adds image understanding while matching the base model's text performance. The sparse mixture-of-experts model uses 13B active parameters out of 284B total and supports a 1M token context window.
Google DeepMind published a retrospective on 15 years of game-based AI research, tracing a line from 2015's Atari-playing DQN through AlphaGo, AlphaZero, MuZero, and AlphaStar to its current generalist agent, SIMA 2. The post also details a new research partnership with Fenris Creations, the studio behind EVE Online, to study continual learning, memory, and long-horizon planning in a persistent multiplayer universe.
OpenAI has released a plugin allowing ChatGPT to read and respond to Apple iMessages on Mac, currently limited to ChatGPT Work and Codex users. The feature requires users to grant Full Disk Access and contact permissions, and arrives amid an active lawsuit between Apple and OpenAI.
Anthropic is revising its 30-day data retention policy after enterprise pushback, allowing regulated-industry customers to store the required data on their own cloud infrastructure instead of Anthropic's servers. The changes, built with more than 100 customers, are set to roll out this fall.
OpenAI is previewing a transparent background feature for GPT-Image-2 through its API, letting developers generate PNGs with no background baked in at generation time. The company claims this produces cleaner results than traditional background removal, particularly on difficult edges like glass or thin fibers.
Cline's v4.1.11 update adds inline image generation for supported models and fixes a code actions bug that broke on VS Code 1.134. The release also expands the model catalog with eight new providers including AMD, Arcee, and RunInfra.
OpenAI has launched an Apple Messages plugin that lets ChatGPT read, sort, draft, and send texts on a user's behalf. The company says the plugin runs locally and does not index all messages, though it has not detailed exactly how that works.
Amazon Bedrock now supports cross-Region inference for three GPT-5.6 variants — Sol, Terra, and Luna — across more than 25 AWS Regions. The feature routes requests to available compute capacity via geographic or global inference profiles, without requiring code changes beyond swapping a model ID.
Google has updated Gboard's Writing tools on the Pixel 11 with a new 'Suggested' tab powered by Gemini Intelligence. The feature drafts personalized message suggestions using typing history and on-screen context, processed via Google's Private AI Compute.
A stealth model called Ox Alpha has appeared on OpenRouter, offering a 1 million token context window at no cost during its preview period. The model's developer remains anonymous, and OpenRouter says it is acting only as a router, not the model's owner or provider.
OpenAI has added an Apple Messages plugin to ChatGPT for Mac, allowing the AI to read, search, and send iMessage, SMS, and RCS conversations. The feature is available across all plans but currently limited to Apple silicon Macs.