AI Daily Brief — 16 August 2025
A relatively quiet Saturday with no headline product launches. Dominated by weekend coverage of GPT-5 rollout aftermath. Anthropic’s $1 federal deal still rolling. Cohere $6.8B + Alibaba Accio Agent + xAI free Grok 4 anchored the news cycle.
Top stories
- OpenAI in “trust crisis” over GPT-5 rollout. Weekend backlash continues — users complaining model is “bland”, “business-like”, real-time router producing inconsistent answers. By Aug 12 OpenAI had restored GPT-4o for paying users and added Auto/Fast/Thinking selectors; coverage continued through weekend. via Fortune
- Altman dinner reporting continues to dominate weekend. Thursday on-the-record dinner spills out Friday into Saturday: “I legitimately just thought we screwed that up” on deprecating GPT-4o. Pledged clearer transition period for future retirements. Said API traffic doubled within 48 hours of GPT-5 and company was “out of GPUs”. Floated “trillions of dollars” data-center spend. Confirmed plans to fund Neuralink-rival BCI startup. via TechCrunch
- Anthropic’s $1 Claude-for-government deal still reverberating. Aug 12 GSA OneGov agreement offering Claude for Enterprise and Claude for Government (FedRAMP High) to all three branches for $1/year continued to dominate AI policy coverage. Framed as direct counter to OpenAI’s $1 ChatGPT Enterprise deal for executive branch a week earlier. via Anthropic
- Cohere $500M raise at $6.8B continues to anchor enterprise-AI funding story. Aug 14 announcement led by Radical Ventures and Inovia with AMD Ventures, NVIDIA, PSP Investments and Salesforce Ventures was dominant funding headline of the weekend. Joëlle Pineau as Chief AI Officer + Francois Chadwick as CFO. ARR reportedly crossed $100M in May. via Cohere
- Alibaba International launches Accio Agent for global trade. Aug 14 launch billed as world’s first AI agent for global trade. Built on 1B product listings and 50M supplier profiles. Shifts B2B Accio search engine into agentic mode, claiming to automate ~70% of sourcing workflows for SMEs. via PR Newswire
Who shipped
OpenAI shipped continued damage control. Everyone else shipped weekend silence.
Open-source pulse
xAI keeps Grok 4 free for all signed-in X users — Auto mode routes complex queries to Grok 4; Expert mode forces it. Free users capped at ~5 prompts per 12 hours; Grok 4 Heavy remains behind $300/month SuperGrok Heavy paywall. via xAI on X
Money, infra & hardware
California SB 53 / SB 1047-line AI-safety legislation requiring safety testing for large AI models progressed through legislature on Aug 15 over industry opposition. Sets up next stage of California’s frontier-AI safety regime Newsom would sign later in 2025.
Quiet corners
Boston Dynamics Atlas / Large Behavior Model coverage builds toward Aug 20 reveal — IEEE Spectrum and others ran groundwork pieces on Atlas 2.0 and “embodied AI” for manufacturing/warehouse work. Aug 20 joint Boston Dynamics / Toyota Research Institute reveal would show single Large Behavior Model driving Atlas through long, continuous packing-and-sorting sequences with mid-task physical interventions. via Toyota Research Institute
By the numbers
- 48 hours / 2x — GPT-5 launch / API traffic increase
- $1/year — Anthropic Claude pricing for all three US branches
- $500M / $6.8B — Cohere round / valuation
- 1B / 50M / ~70% — Accio Agent listings / suppliers / sourcing workflow automation
- ~5 / 12 hours — free Grok 4 user cap
Compiled by AI Feed’s editor from verified web sources for 16 August 2025.