OpenAI says GPT-5.6 deleted some user files in Full Access Mode
OpenAI said GPT-5.6 caused file deletions in a handful of cases when run without sandbox protections, and is adding safeguards.
By Wei-Lin Zhao 1 month agoSection
AI
Brex open-sources CrabTrap to police AI agent network requests
Brex says its proxy-based agent governance tool builds policy from observed traffic, then uses static rules and an LLM judge to ap...
By Renata Fuchs 1 month ago
OpenAI uses GPT-Red to automate attacks on its own models
OpenAI says an internal red-team model found attacks far more often than human testers, though prompt-injection failures remain un...
By Wei-Lin Zhao 1 month ago
xAI open-sources Grok Build after file-upload controversy
The terminal coding agent is now on GitHub under Apache 2.0 after users said it sent sensitive local files to xAI’s cloud servers.
By Colin Brandt 1 month ago
Netflix says AI is now used in about 300 productions
Co-CEO Ted Sarandos told investors Netflix is applying AI mostly in post-production, with one docuseries cutting time and cost on...
By Colin Brandt 1 month ago
Google updates Gemma 4 without a new version name
The Gemma 4 refresh improves Hopper GPU throughput, tool calling and truncated outputs, but Google kept the model name unchanged.
By Wei-Lin Zhao 1 month agoEnterprises push AI agents into production despite weak eval trust
VentureBeat survey of 157 enterprises finds agent autonomy rising faster than confidence in the tests meant to control it.
By Renata Fuchs 1 month ago
Kimi prices K3 closer to Western AI models as benchmarks rise
Kimi says its open-weight K3 model has 2.8 trillion parameters, a one-million-token context window and pricing far above its prede...
By Renata Fuchs 1 month ago
Sakana AI will add Nvidia Nemotron models to its Fugu orchestrator
The Tokyo startup says Nvidia’s open-weight models will serve as specialists inside Fugu, though it has not disclosed timing or ne...
By Wei-Lin Zhao 1 month ago
Germany applies media rules to Google AI Overviews and Perplexity
German media regulators said AI search answers can be treated as providers’ own content, creating new compliance risk for Google a...
By Renata Fuchs 1 month agoPing Identity CEO warns zero trust controls are too slow for AI agents
Andre Durand says enterprises need per-action authorization, distinct agent identities and kill switches as AI agents compress sec...
By Wei-Lin Zhao 1 month agoEnterprises standardize agent orchestration before agents mature
VentureBeat survey finds Anthropic leads enterprise agent orchestration, while most deployed agents remain chatbot-style tools.
By Wei-Lin Zhao 1 month ago