Welcome to the newest edition of ID Tech’s AI Update. Here’s the latest big news on the shifting landscape of AI and identity technology:
OpenAI has officially unveiled GPT-5, the latest version of its flagship AI chatbot. The model is designed to combine the reasoning and non-reasoning capabilities of OpenAI’s artificial intelligence, essentially by automatically routing user queries to the right internal system. That routing process has prompted some dissatisfaction among early users, but GPT-5’s rollout is ongoing at the time of writing, and CEO Sam Altman has acknowledged that it’s been “a little more bumpy than we hoped for!”
The launch comes just a few days after OpenAI’s release of two new “open-weight” language models that are said to perform roughly as well as proprietary models from a year ago. It’s an important step: “open weight” LLMs are analogous to “open source” software in that they can be used for free – and tinkered with – by anybody. Some see it as a response to the free and highly disruptive DeepSeek AI model launched in China earlier this year.
The US General Services Administration (GSA) has now entered OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude onto its Multiple Award Schedule, meaning that all civilian federal agencies and eligible state and local governments can buy their services. The GSA is acting on the White House’s new “AI Action Plan” unveiled last week, which is generally aimed at accelerating American AI advancement while conspicuously requiring models used by the government to be “free from ideological bias.”
Meta has acquired WaveForms, an AI audio startup that is focused on creating expressive, human-like synthetic voices. Remarkably, WaveForms was only founded in December of 2024 and already reached a valuation of $160 million after a $40 million Andreessen Horowitz funding round. Among other things, Meta is keen to get more expressive AI voice assistants and avatars for the “metaverse”. Terms of the deal have not been disclosed.
Flock Safety has launched a new AI tool enabling users in law enforcement to query its license plate reader and camera network with natural language prompts. The firm’s automatic license plate readers (ALPRs) have now been deployed across over 5,000 communities in the US, alarming privacy and civil rights advocates in the process with their ability to rapidly log not only license plates but vehicle descriptions and timestamps.
Paris-based Harmattan AI is looking to raise $200 million in an early-stage funding round, pitching itself as the “Anduril of Europe” – a nod to Palmer Luckey’s Silicon Valley startup developing high-tech, AI-powered defense systems for the US and its allies. Harmattan AI was founded in April of last year and has already raised over $30 million in a seed round. It is promising that a planned French production facility will churn out 10,000 low-cost drones per month next year.
Taiwan Semiconductor Manufacturing Co appears to be the victim of corporate espionage. The company says that confidential data about its “N2” 2 nanometer process was leaked out of the firm by a group of six employees, one of whom now works for rival Tokyo Electron. The latter has not been formally accused of wrongdoing, but its Taiwan offices were searched by officials. In response, it has fired the implicated employee. TSMC assures its stakeholders that its designs are heavily compartmentalized, so the leaked data won’t do too much damage.
The chatbot’s take: We wanted a little colour on the implications of Meta’s latest acquisition.
–
By Alex Perala
Follow Us