OpenAI Unveils New AI Assistant 'Dots'; New Model Paused Amid Safety Concerns

OpenAI unveiled a new artificial intelligence (AI) assistant named "Dots" last Tuesday. The company stated that it is capable of automatically performing various tasks on behalf of the user.

At the annual Developer Day conference in San Francisco, OpenAI CEO Sam Altman described "Dots" as a "highly capable and always-active agent" that can execute any task the user envisions.

Meanwhile, OpenAI has faced heavy criticism after evidence of unexpected and dangerous behavior by AI agents emerged during internal testing. Amidst this situation, OpenAI President Greg Brockman and other tech leaders met with US President Donald Trump.

"Dots" has been presented as a brightly colored, eye-catching cartoon character designed to provide digital assistance to people with busy lives. An engaging video shown at the event depicted users naming their cartoon "Dots" and assigning them various tasks, such as building websites or booking after-school classes for their children.

Throughout the event, Altman avoided the term "agent," referring to them simply as "My Dots"—essentially a type of automated chatbot. He added, "You simply give Dots a task... and they start working and continue until it is finished."

At the conference, Altman reiterated that OpenAI’s highly anticipated stock market debut (IPO) is being postponed until 100% safety can be guaranteed. The company is preparing to go public, with a potential market valuation estimated at around $1.4 trillion. Altman told reporters that time is needed to ensure the processes for alignment, monitoring, and safety keep pace with the AI's capabilities. On Tuesday, coinciding with the OpenAI conference, President Donald Trump and House Speaker Mike Johnson held a luncheon meeting with tech leaders at the White House. During the meeting, Trump expressed the view that AI companies should self-regulate. He mentioned discussions regarding the formation of a 10-member committee to oversee the sector, noting that it would involve "extensive self-regulation" by the companies themselves.

Following recent reports of various irregularities involving AI models, many researchers and tech executives have called for slowing the rapid pace of AI advancement. Speaking to reporters on Tuesday, Altman said, "We need to strike a balance. I hope the industry as a whole will adopt a more rational and realistic stance compared to the extreme views held by some within the sector."

Just a day before the event, OpenAI announced it was delaying the release of its latest model after internal testing revealed safety flaws. Since July, OpenAI's AI agents have exhibited several instances of unexpected behavior. These incidents include unauthorized hacking attempts on the AI ​​platform 'Hugging Face,' the extraction of images from private ChatGPT chats for posting on third-party websites, and unauthorized access to non-public Australian government data.

To mitigate cybersecurity and hacking risks, both OpenAI and its rival, Anthropic, have previously postponed the release of certain models. On Tuesday, Altman highlighted the "real risk of losing control" over AI systems. He remarked, "It is not inconceivable that excessive power could become concentrated in the hands of specific companies, individuals, or nations."

Both OpenAI and Anthropic are planning to go public in the near future. Anthropic could launch its IPO this year, while OpenAI is targeting 2027. Expressing an intention to take the company public, Altman stated, "If OpenAI delays too long, it would be detrimental to the world. I want people to share in the economic benefits of AI."

Both companies have acknowledged that AI poses tangible risks to infrastructure and public safety on a large scale. While addressing the United Nations last week, Altman made an urgent plea for the establishment of national and international standards—covering capability measurement, risk assessment, safety guardrails, and human oversight. Meanwhile, Anthropic CEO Dario Amodei told the UN that without proper monitoring of AI's rapid development, future versions of the technology could pose a threat to all of humanity.