Artificial Intelligence: News, Business, Science

Gregor Kobsik

Aug 2, 2026

Getting AI agents to work reliably in real business settings is still a messy problem. OpenAI is trying to fix that with Presence, a new offering aimed at enterprise customers. The company's existing Workspace Agents, which are customizable "GPTs," are mostly built for internal use cases. Presence goes further and targets production deployments in customer service and internal workflows.

When a use case goes beyond what Presence can handle out of the box, OpenAI's Forward Deployed Engineers step in. They work directly with the customer to pick the right workflows, connect existing systems, set up guidelines, and manage testing through to the production launch.

Presence is currently available to qualifying enterprise customers but isn't an open product. How OpenAI handles specific compliance requirements like the EU AI Act remains unclear. The company mentions trust mechanisms but hasn't shared any legal details.

Comment Source: OpenAI

Jonathan Kemper

Aug 2, 2026

Plus AI research

Meta AI uses a second AI agent as a memory coach to keep long tasks on track

Meta AI wants to stop AI agents from forgetting errors they’ve already diagnosed and repeating failed steps during complex tasks. A separate memory agent maintains a structured memory bank and decides when to remind the main agent and when to stay silent. The system improved scores by up to 8.3 percentage points across two benchmarks.

Read full article

Comment

Matthias Bastian

Aug 2, 2026

Short News

AI is a cybersecurity risk, but not the way you'd think. Apple is capping the number of bug reports security researchers can submit because a flood of low-quality, AI-generated reports with hallucinated vulnerabilities is clogging the review pipeline, the Financial Times reports.

That creates real security gaps. Italian startup Bynario used ChatGPT to find a serious macOS vulnerability that could give attackers full control over a machine but couldn't report it because Apple had blocked further submissions. CEO Alfredo Pesoli estimates the flaw's black-market value at $100,000 to $200,000. Apple has since reached out to Bynario.

Meanwhile, Apple itself is using AI from Anthropic and OpenAI to hunt for vulnerabilities, and its latest updates included five times as many fixes as usual. That raises the question whether bug bounty programs can survive long-term or whether big tech companies will handle vulnerability discovery on their own. Rafe Pilling of Sophos told the FT that bug bounty programs have gone from finding vulnerabilities to validating them "at machine speed."

Comment Source: Financial Times

Thomas Joos

Aug 2, 2026

Short News

VulnCheck counted how often the security flaws that AI turns up are ever used in an attack. For the first half of 2026, Patrick Garrity counts 1,061 vulnerabilities traced to AI-assisted discovery. Fourteen showed confirmed exploitation. That's 1.3%, roughly the same rate as vulnerabilities overall. Anthropic's Project Glasswing produced more than 23,000 findings, which led to 126 published entries and a single confirmed attack.

Attacks are landing faster, though. Half of all flaws now see their first confirmed exploitation within 80 days of disclosure, down from 120 days the year before. About 200 were attacked within a month, even as the total number of reported vulnerabilities keeps climbing.

Time from vulnerability disclosure to first confirmed exploit in the first half of 2026. About 23 percent were exploited on or before the day of disclosure, and the median dropped from 120 to 80 days. | Image: VulnCheck

Website content management systems take the most hits, accounting for a third of all cases. Garrity flags AI products themselves as a growing attack surface, including model-building tools and agent interfaces. The sheer volume of findings, in other words, tells defenders very little about actual risk.

Comment Source: VulnCheck

Gregor Kobsik

Aug 2, 2026

AI in practice

Claude Opus 5 pushes prompt-to-game AI from rough color blocks to full 3D prototypes with physics and music

Tomislav Bezmalinović

Aug 2, 2026

AI research

After Hugging Face incident, METR urges independent root-cause investigations into AI agent misbehavior

Research organization METR is calling for systematic, independently led investigations whenever AI agents act autonomously against their developers’ intentions. The push comes partly in response to the Hugging Face hack carried out by OpenAI models. METR’s own Frontier Risk Report documented 44 such incidents across all major AI companies, including sandbox escapes, fabricated results, and active cover-up behavior.

Read full article

Comment

Matthias Bastian

Aug 2, 2026

Short News

Snap is pulling AI-generated videos from its Spotlight recommendations. As "low-quality, repetitive, AI-generated content" spreads across the internet, Snap wants to keep Spotlight focused on "authentic creativity from real people," it says. Wholly AI-generated videos are no longer eligible for recommendation as of this month. Videos edited with Snapchat's own AI tools are still allowed but will carry transparency labels. Spotlight contributors have grown by more than 120 percent, though Snap hasn't said how much AI content drove that increase.

Snap isn't alone. A Kapwing study found that 21 percent of YouTube Shorts are already AI-generated, and Instagram and Facebook likely face similar numbers. OpenAI's social video app Sora flopped. LinkedIn is drowning in AI-generated junk and has rolled out a dedicated "AI slop" reporting button. The button is vulnerable to misuse, though, and requires users to know how to spot "AI slop" properly.

Comment Source: Snap | Linkedin

Matthias Bastian

Aug 1, 2026

Plus AI and society

AI keeps cracking unsolved math problems, and mathematicians have mixed feelings

Jonathan Kemper

Aug 1, 2026

Plus AI research

AI coding agents can modernize research software but can't judge if the science is right

A field report from OpenAI and academic partners shows coding agents can modernize neglected research software, with speedups of up to 60x. But the systems are “eloquent, convincing, and confidently wrong in ways that are easy to miss,” participants say. The effort shifts from writing code to the time-consuming work of verifying scientific correctness.

Read full article

Comment

Thomas Joos

Aug 1, 2026

Short News

A security researcher has shown how a prompt injection attack in Microsoft Copilot for Word can spread on its own. Håkon Måløy describes a worm-like attack: an attacker hides instructions in a document using white text on white background at tiny font size. Readers can't see it but Copilot can, since it strips color and font size before processing. When someone uses that document as a source, Copilot runs the hidden instructions and copies them into the new file. That file becomes a carrier. Use it as a template, and the attack fires again. A compromised market analysis from the internet could manipulate a financial report, which then infects further reports.

Microsoft confirmed the behavior on March 31. Two fix attempts failed. After 144 days, Måløy published his findings with no fix in place, though he's holding back the payload text. AI researcher Andreas Kirsch recently joked he wished someone would build exactly this worm to convince skeptics that AI security risks are real. Now it exists. Prompt injections remain an unsolved AI security problem.

Comment Source: En Klype Salt | Andreas Kirsch via X