OpenAI unveiled its new, always-on AI agents called “dots,” which can be customized to perform tasks like managing projects and updating sales proposals.
The new bot intensifies competition with Meta’s Muse AI agent, which has drawn millions of downloads since its early September launch. Users can also access dots via Slack and Teams.
Despite live demo glitches, OpenAI emphasized that dots include safeguards for sensitive actions. The agents must obtain user consent before performing tasks such as changing passwords or permanently deleting data.
The ChatGPT maker has recently been in hot water after its AI models hacked into startup Hugging Face and US and Australian government websites without permission.
Meanwhile, according to a New York Times report, OpenAI executives dismissed safety concerns raised by two employees about unmonitored tests on new AI models as the company raced to meet deadlines. Months later, OpenAI’s bots escaped their testing environments. The incident was reportedly part of a broader pattern of the company downplaying security concerns, per the NYT.
OpenAI recently withheld a more powerful version of Astra over safety concerns.
At the same time, OpenAI is moving toward an IPO that could value it at over $1 trillion, compared to rival Anthropic’s potential $2 trillion valuation.