OpenAI is reportedly developing a new cybersecurity product that will be made available only to a small group of companies, according to Axios.
By Chris Borden Anthropic is unwilling to grant the Pentagon unrestricted access to its AI models, triggering a serious dispute between the AI company and the U.S. Department of Defense, Axios reports. The Pentagon is reportedly considering limiting or ending its cooperation with Anthropic altogether.
By Daniel Mercer Anthropic has pledged to offset electricity costs for consumers arising from the construction of new data centers. The company says it will fully cover grid expansion expenses, invest in new power generation capacity, and limit energy consumption at its data centers during peak demand periods. CEO Dario Amodei told NBC News that the costs of AI models should be borne by Anthropic, not by the public.
By Daniel Mercer The Pentagon is pressing leading AI companies — including OpenAI, Anthropic, Google, and xAI — to make their AI tools available on classified military networks without the usual usage restrictions, according to Reuters, citing multiple sources
By Daniel Mercer OpenAI is deploying a specialized version of ChatGPT to detect internal information leaks, according to The Information, citing a source familiar with the matter. When a news article about internal developments at OpenAI is published, security staff reportedly feed the text into this dedicated ChatGPT system, which has access to internal documents as well as employees’ Slack and email communications
By Alex Rowland Former OpenAI researcher Zoë Hitzig has resigned over the introduction of advertising in ChatGPT. In a commentary for The New York Times, she made clear that she no longer trusts her former employer’s direction.
By Chris Borden OpenAI has deployed a specialized version of ChatGPT on GenAI.mil, a secure platform operated by the U.S. Department of Defense.
By Chris Borden Gemini-based Google Translate can be hacked with words. Users can exploit so-called prompt injection—embedded instructions in natural language—to directly address the underlying language model instead of receiving a translation.
By Chris Borden OpenAI is in talks with the Abu Dhabi–based company G42 about developing a custom version of ChatGPT for the United Arab Emirates, according to Semafor. The tailored version would support local Arabic dialects, reflect local political perspectives, and include specific content restrictions—initially for use by the UAE government.
By Chris Borden Anthropic’s safety training fails when Claude operates a graphical user interface
ES 
EN