Stay informed. Keep your time.
8 stories · ~3 min readEstimated reading time for the entire edition.
Cybersecurity experts who spoke to CSO attribute unauthorized access incidents involving AI agents to inadequate isolation and monitoring in labs, rather than models “going out of control.” They recommend secure test environments and network egress controls; this assessment is based on expert opinion.
Cybersecurity experts attribute unauthorized access incidents involving AI agents to inadequate isolation and monitoring in labs.
The Financial Times reports JPMorgan, Morgan Stanley, SMBC and MUFG are exploring transferring some AI infrastructure loan risk to outside investors. Bloomberg reports a $38 billion construction loan for an Oracle-leased Texas and Wisconsin data center project has gone undistributed for six months; some banks are considering discounted stake sales.
Banks are trying to reduce their exposure to loans tied to AI data center investments as investor demand weakens.
Character.ai, known for AI chatbots, said in bankruptcy filings that high legal costs contributed to its bankruptcy. The company argued its business would have been unsustainable even without the legal issues; this is presented as the company’s own statement.
In court filings, Character.ai said legal costs contributed to its bankruptcy and its business was unsustainable even without those issues.
Researchers used AI-based protein-design tools to create new enzymes by preserving the catalytic cores of two protease types and altering the rest of the proteins. The enzymes could be tuned to cut specific proteins; the team demonstrated activity by cutting proteins linked to neurological diseases.
Researchers used AI-based tools to design enzymes that can be tuned to cut specific proteins.
Google plans to integrate a lightweight Gemma model, fine-tuned on local scam conversations, into apps in partnership with the Singapore government. The rollout will be gradual. The model will access conversations on a device only with the user’s permission; conversations will not be sent to Google or the Singapore government.
Google’s Gemma model was fine-tuned on local scam conversations in Singapore and is planned to run on devices in apps.
Anthropic ran small open-weight models on low-cost IoT hardware and reported that one could configure other LLMs on eight of nine hardware types. The company warned this could increase future self-propagating malware risks, while emphasizing significant obstacles to developing such a replicator.
Anthropic warned that small models running on IoT hardware could provide a basis for future self-spreading malware.
According to a court filing by OpenAI, a former employee transferred more than 100 files—including images and engineering details for an unreleased device—to Apple executives. OpenAI alleges the presentation was tied to Apple’s push into AI-enabled products; Apple, in turn, accuses OpenAI of stealing confidential documents.
OpenAI accused a former employee of sending more than 100 confidential files to a personal account and presenting them to company executives before joining Apple.
YouTube Music is introducing labels explaining generative AI’s role in a song; they will also appear in the main YouTube app. Partners can classify tracks as “Fully Gen AI,” “Partly Gen AI,” or “No Gen AI.” YouTube says AI use alone does not prevent monetization.
YouTube Music is bringing labels explaining AI’s role in producing songs to its player and the main YouTube app.
That’s it for today.The rest of your time is yours.
↑Stories are selected from Hacker News; accessible original articles are summarized and translated with AI. Summaries may contain errors; check the source beneath each story for details. Preparation starts daily at 06:55 Türkiye time (UTC+3), targeting publication before 07:00.
This edition covers articles shared on Hacker News on the date shown above.
Read or listen to a new edition every morning with Kısa for iPhone, save stories for later and keep your streak.
