August closed out with one other busy week for AI developments. OpenAI positioned new guardrails on its Astra mannequin, the U.S. authorities weighed in on “truthful use” copyright legal guidelines for AI fashions and Nvidia plans to amass an open-weight mannequin firm.
Learn on for a take a look at a few of the week’s prime tales and what they imply for CIOs.
Information Highlights Aug. 31–Sept. 4
OpenAI tightens the reins on Astra mannequin
What occurred: OpenAI is inserting restrictions on its Astra mannequin and including extra safety safeguards after figuring out that the AI mannequin is a cybersecurity threat. The corporate is limiting the mannequin’s entry to cybersecurity capabilities. It has been skilled to refuse requests for malicious actions and determine jailbreak makes an attempt. OpenAI mentioned that whereas Astra wasn’t the mannequin concerned within the Hugging Face breach, the corporate utilized learnings from that incident towards enhancing the safety of its fashions.
Why it issues: It is critically vital for CIOs and CISOs to guide the cost in making use of applicable guardrails and acceptable use insurance policies for AI. Human-in-the-loop insurance policies, AI governance councils and inner coaching are among the many initiatives CIOs and CISOs can make use of to make sure AI does not exit the sandbox to enact cyber threats.
Trump administration sides with OpenAI on ‘truthful use’ debate
What occurred: A letter from the Division of Justice defined that the Trump administration is backing OpenAI within the AI firm’s authorized battle with The New York Occasions and different publishers. The federal government claims that coaching AI on information content material is appropriate in keeping with “truthful use” copyright legal guidelines and that limiting truthful use might negatively have an effect on competitors for big language fashions (LLMs).
Why it issues: The content material LLMs are skilled on — and the content material AI generates — has lengthy sparked copyright legislation debates. Lawsuits in opposition to AI firms, together with OpenAI and Meta, have been filed relating to the coaching of fashions on copyrighted content material. CIOs have threat mitigation instruments at their disposal to handle the copyrightability of AI outputs, together with obligatory disclosure necessities that require staff to doc AI-generated content material. CIOs can even audit the supply of coaching information when creating their very own inner AI fashions.
Nvidia to amass Hugging Face for $13 billion
What occurred: AI chipmaker Nvidia plans to buy AI startup Hugging Face for $13 billion. Hugging Face gives open-weight fashions and an open ecosystem by which anybody can add and take a look at AI fashions. Nvidia CEO Jensen Huang informed The Wall Avenue Journal that open fashions present organizations with the power to “construct on superior capabilities with out coaching each mannequin from scratch.”
Why it issues: In distinction to propriety fashions by firms together with OpenAI and Anthropic, open-weight fashions are public, democratizing entry to AI fashions and offering CIOs and IT groups with the chance to obtain and use these fashions in their very own environments. There are a number of vital downsides that CIOs must weigh when contemplating the usage of open-weight fashions. In a current weblog, Anthropic identified that open-weight fashions “probably current the next threat than closed fashions.” Self-hosting AI fashions can even require costly GPUs and generate excessive upkeep prices.
Google’s new AI mannequin
What occurred: Google is growing a brand new AI mannequin, Gemini 3.8 Flash, with up to date coding capabilities, which might assist the corporate catch as much as rivals akin to OpenAI and Anthropic. Of late, “Gemini has returned to trailing flagship fashions from Anthropic and OpenAI at a second when agentic coding has taken off because the premier enterprise use case for AI,” reported The Wall Avenue Journal.
Why it issues: If Google can catch as much as frontier mannequin rivals, CIOs have extra choices to select from as they put money into AI fashions. Nonetheless, many CIOs are working towards integrating their AI investments this 12 months to stability the event of inner AI applied sciences with deployments from third-party distributors. Having extra AI mannequin selections can profit pricing however can result in determination paralysis if a company’s total AI technique stays unclear.
From InformationWeek: Definitely worth the learn
You may discover a wealth of perception and evaluation on InformationWeek, however listed here are just a few curated picks from the week that needs to be on the prime of your studying checklist.
Intuit, Smartsheet, ETS CISOs on making certain enterprise resilience earlier than ‘AI orphans’ emerge
With out guardrails in place, the proliferation of AI brokers inside organizations can create complications round identification administration and go away “orphaned” AI brokers with unfettered entry to non-public information. CISOs clarify how identification administration is turning into extra complicated and the next safety threat with the rise of AI brokers within the enterprise.
InformationWeek Podcast: Overcoming center managers’ AI pushback
Aleks Bass, CPTO at Typeform, and Kendrick Kho, CAIO at Scrubbed, be part of Senior Editor Joao-Pierre Ruth on the InformationWeek Podcast to debate approaches to profitable over reluctant group leaders who elevate issues about AI.
AI, management, {hardware}: 10 IT tales that outlined summer time 2026
AI breakthroughs, IT management challenges, {hardware} complications — CIOs had a lot to take care of this summer time. Managing Editor Jacqui Biscobing examines 10 InformationWeek tales that captured your consideration.
The following technology of CIOs will take a unique path to the highest
Three CIOs clarify how AI is altering their jobs and the abilities and backgrounds the position requires now. One high quality stays fixed: curiosity.
Arising: Dates to observe
Sept. 7-11 — Subsequent week, Oracle, FedEx, Macy’s and Adobe host earnings calls with monetary analysts.
Sept. 15-17 — Salesforce’s Dreamforce occasion is coming to San Francisco this month. Dario Amodei, CEO of Anthropic, is scheduled to talk along with audio system from MIT, the Raspberry Pi Basis, Time, Black Forest Labs and extra.
Sept. 23-25 — WeAreDevelopers World Congress can be held in San Jose, Calif., bringing collectively greater than 10,000 builders, 4,000 firms and 500 audio system. Periods will cowl subjects akin to coding AI brokers, establishing guardrails for brokers, programming languages and extra.
This week’s AI information reiterates the significance for CIOs to set guardrails and acceptable use insurance policies round their AI deployments. We would love to listen to from you — how is your group making certain that AI operates securely throughout the enterprise? Tell us at [email protected].
