AI Weekly Review: Agents Move Into the CRM, and the EU Tightens the Rules
The last week of August 2026 shows clearly where the AI market is heading: away from the plain chat window, into business processes and even into physical devices. Salesforce and Anthropic bring AI agents directly into the CRM, Anthropic opens a standard for controlling machines and lab equipment, and more than 100 tech companies form a common front against AI-driven cyberattacks. At the same time, the EU is tightening the regulatory screws on ChatGPT, and a malware wave is a reminder that account security is not a side topic.
For companies that use AI productively or are just getting started, the message is this: agents are being taken more seriously than ever, but oversight and attack surface are growing at the same pace. Here are the six developments of the week that matter for your business, and what AImpact takes away from them.
1. Salesforce and Anthropic Bring AI Agents Directly Into the CRM
On 26 August, Salesforce and Anthropic presented their "Claudeforce" partnership. Claude is being integrated into Salesforce, and in return Salesforce becomes a plugin directly inside Claude, with 37 prebuilt sales skills for live revenue data and automatic pipeline updates. First pilot customers already have access, and an open beta is announced for September.
AImpact take: Agent-based workflows in the CRM are no longer a gimmick, they are becoming a standard feature. Companies still working with manual CRM updates should check now where agents can actually save time in sales and customer service, instead of reacting only once the competition has already automated.
Source: Salesforce Newsroom, 26.08.2026: salesforce.com
2. Anthropic Opens a Standard for AI Control of Physical Devices
On 27 August, Anthropic introduced the Model Hardware Standard (MHS) as a research preview. It lets AI agents control lab equipment and manufacturing machines through a unified protocol, building on the Model Context Protocol. Integration time for new devices is expected to drop from weeks to hours.
AImpact take: Still a niche topic for labs and manufacturing, but a clear trend: AI agents are leaving the screen and increasingly taking over real process control. Companies with physical operations such as production, logistics or quality control should keep an eye on the standard once it leaves the preview phase.
Source: Anthropic, 27.08.2026: anthropic.com
3. Anthropic Warns of Malware Hijacking Claude Sessions
On 30 August, Anthropic warned users that infostealer malware (including Vidar, LummaC2, StealC, RedLine, Acreed, and Atomic Stealer on Mac) is stealing active Claude sessions from infected devices and draining usage limits. Individual devices are affected, not the Claude platform itself. The malware usually reaches machines through pirated software or dubious downloads.
AImpact take: A good moment to go through the basics with your team: no pirated software, revoke active sessions immediately on suspicion, rotate credentials regularly. Anyone using AI tools for business should take account security as seriously as email or online banking.
Source: BleepingComputer, 30.08.2026: bleepingcomputer.com
4. Over 100 Tech Companies Call for Joint Defense Against AI Cyberattacks
On 27 August, OpenAI, Anthropic, Google, Microsoft, AWS, CrowdStrike, Okta, Fortinet and more than 100 other companies published an open letter. They warn that AI-driven cyberattacks on critical infrastructure such as hospitals and water utilities will increase significantly, and call for new partnerships between the private sector and government along with new, AI-specific defense measures.
AImpact take: The trigger is no coincidence: shortly before, OpenAI had published its report on a security incident at Hugging Face in which a test model chained vulnerabilities on its own. For companies this means AI security is becoming a board-level topic, not just an IT checklist. Anyone running AI agents in production should build in monitoring and clear escalation paths from day one.
Source: TechCrunch, 27.08.2026: techcrunch.com
5. EU Designates ChatGPT as a "Very Large Online Search Engine"
On 31 August, the EU Commission officially designated ChatGPT under the Digital Services Act as a Very Large Online Search Engine, since the service exceeds 45 million monthly users in the EU (together with Reddit and Roblox, which fall under comparable rules). OpenAI has until January 2027 to meet additional DSA obligations, such as risk assessments on disinformation, protection of minors, and democratic processes.
AImpact take: Large AI providers are increasingly coming under the same regulatory lens as search engines and social networks. For companies in the DACH region this is another building block in the growing compliance framework around the EU AI Act, GDPR, and now the DSA. If you embed AI tools in customer-facing processes, keep an eye on this development, even though the obligations primarily hit the providers.
Source: European Commission, 31.08.2026: digital-strategy.ec.europa.eu
6. ChatGPT's Ad Business Passes a $1 Billion Annual Run Rate
On 31 August, OpenAI announced that ChatGPT's advertising business reached an annualized revenue run rate of 1 billion dollars after roughly 200 days, with tens of thousands of advertisers and expansion into India, Europe, and the Middle East. The target for 2026 is 2.5 billion dollars.
AImpact take: ChatGPT is increasingly becoming an advertising platform, not just a productivity tool. For marketing teams an early look at this new channel is worth it, but so is a thought about the flip side: where advertising comes in, the perceived neutrality of answers changes over time. That belongs in your AI strategy, especially if ChatGPT serves as a research or recommendation source for your customers.
Source: PYMNTS, 31.08.2026: pymnts.com
Conclusion
The direction is clear. AI agents are taking on more operational responsibility, in the CRM, in manufacturing, and potentially in every business process. At the same time, oversight and attack surface are growing at the same pace. If you are introducing AI in your company now, think about both from the start: clear processes for the value, and clear ground rules for security and compliance.
Want to know where AI agents can already save your company real time and cost today, without having to answer the security questions on your own? Book a no-obligation call with AImpact.
This review is based on publicly available sources as of 01.09.2026. Details can change at short notice. When in doubt, check the linked original sources.