OpenAI debuts 'Dots' AI agent at DevDay, Valley shops weigh risks
OpenAI CEO Sam Altman introduced an 'always-on' agent called Dots at a San Francisco developer event, a day after pausing a separate model over safety concerns. Central Valley governments, schools and vendors say the security questions land close to home.
OpenAI debuts 'Dots' AI agent at DevDay, Valley shops weigh risks
Key Takeaways
- OpenAI unveiled "Dots," an always-on AI agent, at its San Francisco DevDay on Sept. 29, 2026.
- CEO Sam Altman's keynote skipped detailed security concerns tied to a paused model, per AP reporting.
- Fresno County's CalSAWS contact center is already testing generative AI post-call summaries.
- UC Merced and Fresno State groups have AI efforts under way, making agent safety a live question.
Onstage in San Francisco on Tuesday, OpenAI CEO Sam Altman pitched a new "always‑on" assistant called Dots. Local governments and vendors in the Central Valley that already lean on AI helpers for contact centers and campus projects are watching, because agent security turns from blog talk into real work here.
The name is Dots.
What OpenAI announced
Altman’s keynote walked developers through tools meant to let an agent work across apps, keep running without a prompt, and hand off tasks between skills. The company framed it as a way to move past chat boxes and into software that does jobs on its own. AP’s account notes Altman didn’t dwell on recent security worries during the onstage remarks, though he fielded questions about safety later.
OpenAI said the agent can remember goals, take actions through connected services, and loop a human in when it’s unsure. That’s the sales pitch. How it behaves in the wild is what will matter to IT directors.
Why this matters in the Valley
Fresno County’s social services contact center, part of the statewide CalSAWS system, has been piloting generative AI for post‑call summaries. If counties eventually let an agent act on cases or tickets, the guardrails around tools like Dots will decide whether that’s safe or sloppy. UC Merced’s CITRIS-affiliated efforts and Fresno State’s AI task force give the region a bench of researchers and admins who will stress test any rollout. Local firms pitching multi‑agent automation to Valley clients are taking inquiries too, even if buyers are cautious.
In plain terms, a persistent agent could triage service requests overnight, draft work orders, update records, then ask a human to approve changes at 8 a.m. That’s where the time savings sits for public agencies and clinics from Merced to Tulare. It’s also where risk lives if the system reaches further than intended.
What wasn’t answered
The launch came a day after OpenAI paused a separate, more advanced model over safety concerns raised in testing. Altman didn’t go deep on that in the keynote. The company has said it’s spending more on monitoring and security for agents, and those promises will meet procurement checklists fast in Fresno and Stanislaus. And Valley IT teams will ask for logs, fail‑safes and clear default permissions before anything touches production.
Developers here will also want to know how Dots handles identity, how actions are scoped inside a county or campus network, and how the tool behaves under load. They’ve seen chat systems go sideways, they won’t accept guesswork from an always‑on bot. One more note from the floor: pricing and service‑level details decide adoption almost as much as features do, so those numbers matter.
By 4 p.m. in Fresno, the air read 94.
Central Valley AI is produced by the CVAI Newsdesk team and developed by Kaweah Tech, a regional firm that builds, deploys, and integrates AI solutions for businesses across California's Central Valley.