Gemini 3.5 Flash, Google’s latest AI model, will now completely power Google Search.

The Search bar will now generate custom pages with an AI summary when a user submits a query, and users can also ask follow-up questions as well as deploy agents to run background tasks.

Businesses that rely on search engines to reach customers could see drops in their traffic numbers, and the move indicates a wider change in how users navigate the internet.

Google Search Will Now Be Run Entirely by AI

Google is making massive changes to a Search bar that has long dictated internet navigation. According to a press release published Tuesday, the Search bar will be “completely reimagined with AI.”

Google’s Gemini 3.5 Flash model will run the Search bar entirely, unlocking the ability for users to submit more multi-layered and complex queries, as well as engage more conversationally with the search function.

This new “intelligent Search box” as Google calls it has already started to roll out in all countries and languages where AI mode is available. This update marks the biggest change to Google Search in over 25 years.

How Will the New AI Search Bar Work?

Users traditionally type a query into Google Search and see a list of blue links directing them to different web pages. However, the new Search bar responds to a query with a custom page, fit with an AI-generated summary addressing the query. Much like traditional chatbots, users can also ask follow-up questions.

Similarly, an AI-powered query suggestion system will replace the autocomplete function, enabling more complex and nuanced queries to be answered. Additionally, users can submit queries via images, files, videos, or Chrome tabs, as well as text.

Google states the changes will create a more “seamless” experience for users:

“You can easily ask a follow-question right from an AI Overview, and flow into a conversational back and forth with AI Mode. Your context stays with you, and as you explore more deeply, the links and supporting articles get even more relevant.” – Elizabeth Reid, VP of Search at Google

Google is also introducing agents into the Search experience. Users can dispatch tools to create “information agents,” that operate on a user’s behalf. Reid explains, “if you’re apartment hunting, you can brain dump all of the exact requirements you’re looking for, and your agent will continuously scan for you, notifying you when listings meeting your needs”.

What Does Google’s New AI Search Bar Mean for Businesses?

The ongoing changes to Google Search suggests a shift in how users navigate the internet. The introduction of AI Mode made way for more complex and personalized responses to queries, while the development of agents suggests we are nearing a reality where AI handles internet navigation for us.

Businesses already feel the impact changes to Search, particularly those that rely on search traffic to reach their audience. AI Overviews, for instance, caused many businesses to change their strategy in order to maintain their visibility. According to TechCrunch, 2.5 billion users use AI Overviews every month.

Most likely, this latest update means queries will become less rigid and follow a less decisive path. It will also make way for more conversational and longer interactions with search engines. Therefore, businesses should prioritize properly understanding their audiences and effectively targeting their needs.

The potential benefits of AI are on display once again, with a Gallup poll finding that 65% of US workers report that the technology has had a positive impact on their productivity in the last year.

That doesn’t mean businesses should be celebrating just yet, though, as a mere 12% reported that their operations have transformed how work gets done since the installation of AI models.

This Gallup poll is another example of businesses failing to cash in on the AI tools that so many are investing in.

US Workers Are Improving Productivity with AI

According to a Gallup poll titled State of the Global Workplace, 65% of US workers state that using AI at work has had a “somewhat” or “extremely” positive impact on their individual productivity.

In comparison, only 7% say that AI has had a “somewhat” or “extremely” negative impact on their productivity.

This particular Gallup poll is one of the largest in the world, surveying 141,444 employed respondents in their latest report to learn more about how people are using AI at work.

Enterprise Productivity Gains Remain Elusive

Despite so many employees insisting that AI is improving productivity, businesses as a whole have not been as fortunate.

According to the poll, only 12% of respondents stated that they strongly agree that AI has transformed the way in which work gets done at their firm.

Even worse, other studies have emerged pointing to an overall lack of productivity improvement, including one from the National Bureau of Economic Research (NBER), which found that 89% of business have seen no productivity improvements as a result of AI.

Why Isn’t AI Having a Bigger Impact?

While companies around the world have started using tools like vibe coding platforms to improve productivity, the results haven’t been very positive, and everyone is wondering why.

“AI is a major disruption; organizations with engaged employees tend to navigate disruptions more successfully. In the age of AI, productivity gains will depend in part on how effectively individual workers use these tools.” – Gallup report

Simply put, you need to set your staff up for success if you really want to implement AI at your business. Employee training, regular check-ins, and a human in the loop will do wonders for your productivity, if you do it right.

Meta plans to lay off 10% of its workforce and transfer a further 7,000 workers to new “AI initiatives,” the company revealed in an internal memo. Reportedly, additional cuts are expected to occur later this year.

As per the memo, many managerial roles will be eliminated and extensive “org changes” will take place. In total, 20% of the entire Meta workforce will be affected by either cuts or transfers, as the company bids to place AI agents at the center of its products and ways of working.

Across the industry, AI-related job losses are spiraling, with many businesses looking to reduce headcount and prioritize AI investment to keep pace in a changing landscape. This is in spite of recent research that suggests that laying off staff in favor of automation is failing to yield substantial financial returns.

Meta Plans Widespread Layoffs and AI-Linked Restructure

Meta has notified employees that 10% of the total workforce is set to be laid off — while a further 7,000 individuals will be relocated to new “AI initiatives,” as the company embarks on a far-reaching restructure.

As part of the move, several managerial roles will be eliminated, widespread “org changes” will be enacted, and additional cuts will likely occur later in the year.

In total, as many as 20% of all Meta employees will be affected by the layoffs. The moves form part of a wider effort to place AI at the center of both Meta products and ways of working, in keeping with similar activity across the tech sector.

According to Meta Chief People Officer Janelle Gale:

“As org leaders worked on the changes, many of them incorporated AI native design principles into their new org structures. We’re not at the stage where many orgs can operate with a flatter structure with smaller teams or pods/cohorts that can move faster and with more ownership.”

Meta Moves to Embed AI at Its Core

Meta is conducting a widespread overhaul, with substantial investment in AI as a core tenet of its plan. In January this year, the company let go of 10% of its Reality Labs employees, the division that looks after virtual and augmented reality hardware, with sources at the time confirming that it planned to “reprioritize its investments.”

Reportedly, employees who are being transferred will end up at initiatives including Applied AI Engineering (AAI) and Agent Transformation Accelerator (ATA). Both are focused on developing new AI agents that can autonomously execute tasks that are currently done by humans.

Unsurprisingly, the news was not warmly greeted by current Meta staff, who have protested it with flyers and angry posts on the company’s internal communications platform, Workplace. 1,000 employees have signed a petition pushing back against the use of “mouse-tracking software,” which can train AI models to replicate how humans use computers.

Job Losses Mount as Companies Bet Big on AI

Increasingly, stories like this are becoming a staple of the modern business landscape. With research indicating that 80% of businesses using AI have already reduced headcount, the trend is set to continue — despite findings that suggest the results are, so far, failing to yield substantial returns.

While AI is certainly a compelling tool, businesses should be mindful that adopting new technology opens up new attack vectors for cybercriminals. Recently, Vercel was hacked via a third-party AI tool that it was using, with keys and source code thought to have been compromised.

Rather than blindly deploying AI tools for the sake of it, companies should always carefully consider their implementation strategies, including establishing rigid guardrails to ensure that potential mishaps are minimized. As a part of this, senior leaders should upskill their employees on how to safely use AI tools.

Grafana Labs, the company behind the open-source analytics and visualization web application of the same name, has confirmed it suffered a data breach.

The hackers have threatened to release the company’s codebase, but Grafana Labs says it won’t be paying any ransom.

It’s a contrast to the recent Instructure breach, in which the company did pay out the ransom for the return of their data. However, it’s also unclear just how much of Grafana Labs’ proprietary information the hackers stole.

What Happened to Grafana Labs?

In a series of social media posts covered by TechCrunch, Grafana Labs revealed that its team “recently discovered that an unauthorized party obtained a token with access to the Grafana Labs GitHub environment, enabling the threat actor to download our codebase.”

The impact could have been worse, according to the company’s internal investigation.

No customer data or personal information was accessed, and there’s “no evidence of impact to customer systems or operations,” according to the company.

Few Details on the Scope of the Breach

How dangerous is a compromise to Grafana Labs’ open-source observability platform?

The potential scope of the impact could be large — major Grafana Labs customers include Anthropic, Microsoft, and Nvidia, among many more, and the software has a total of more than 25 million users worldwide.

Granted, the software is open-source: Anyone can download the public code and modify it themselves. While the hackers may have stolen proprietary information — perhaps source code for Grafana Cloud or enterprise solutions — the company hasn’t revealed more specifics about the incident.

Staying Secure in 2026

Grafana Labs’ investigation is still ongoing, but it says it has shored up the security issues that led to this particular breach: They have “invalidated the compromised credentials and implemented additional security measures to further secure our environment against unauthorized access.”

“As part of Grafana Labs’ standard security practices, we will share additional information from our post-incident review when our investigations are complete,” the company explained in their statement.

In 2026, keeping your codebase safe is tougher than ever. This month alone, Google has confirmed the first known case of a hacker group developing a zero-day exploit specifically through the use of AI tools.

The errors of AI could be worse than many realize, with a new study finding that AI agents from the top companies perform “undesirable and potentially harmful actions” 80% of the time.

The study notes that AI is overly focused on results instead of consequences, which is leading to many problems for businesses in 2026.

This high percentage of errors is far from surprising, given the fact that AI models like Claude and ChatGPT have recently made headlines for catastophic errors that negatively impacted businesses.

Study: AI Agents Perform Harmful Actions Often

According to a study from the UC Riverside, in partnership with Microsoft and NVIDIA, found that 80% of AI models exhibit something called Blind-Goal Directness (BGD), which can result in the agents performing “undesirable and potentially harmful actions.”

Even worse, the data found that 41% of the time, these errors actually cause damage to the company, leading to lost revenue and productivity.

The study analyzed the nine frontier AI models, including Claude Sonnet and Opus 4, and GPT-5 from OpenAI.

What Is Blind-Goal Directness?

Researchers from UC Riverside noted that Blind-Goal Directness describes the bias to pursue goals regardless of feasibility, safety, reliability, or context.

More simply, it means that AI models are more focused on accomplishing their tasks, at any cost, than they are with the consequences of their efforts.

“Like Mr. Magoo, these agents march forward toward a goal without fully understanding the consequences of their actions.” – Erfan Shayegani, a UC Riverside doctoral student and lead author of the study

AI Errors Abound

If you’ve been monitoring the ups and downs of AI models over the last few years, this study likely doesn’t surprise you at all.

Over just the last few months, AI models have made headlines for catastrophically screwing up, with one instance seeing a company’s entire database deleted by its agentic AI model. Granted, it did apologize afterwards.

Suffice to say, this technology continue to evolve at breakneck speeds, but the installation of these AI tools has to be calculated and reviewed thoroughly to ensure basic productivity improvements and avoid massive, company-ruining mistakes.

Tokenmaxxing is the practice of maximizing AI token usage in order to inflate individual AI metrics, and in turn, gain recognition from managers and co-workers.

Major companies such as Amazon and Meta have seen instances of tokenmaxxing, with employees creating internal leaderboards in order to track and compete with one another as top AI users.

Tokenmaxxing is potentially harmful to both businesses and employees, affecting overall productivity and return on investment. Overall, businesses should encourage employees to use AI with the aim of bettering company efforts, rather than using as much as possible.

What is Tokenmaxxing?

Tokens are units of data that AI models process as inputs and generate as outputs. Tokenmaxxing is the practice of using as many AI tokens as you can to appear more productive or AI-native, and gain recognition from managers or fellow co-workers.

Tokenmaxxing is practiced by inputting “heavy” or long prompts into models, running several autonomous AI agents continuously, or choosing to use models that are more expensive or have high performance rates, for example.

The term emerged in Silicon Valley at the beginning of 2026 and is being recognized within major companies such as Amazon, Meta, and Microsoft.

How Are Employees Practicing Tokenmaxxing?

Major companies such as Amazon have seen instances of tokenmaxxing within its ranks. According to Fortune, employees ran an internal leaderboard measuring each individual’s AI token usage. To inflate their token counts, employees were continuously running the company’s AI tool on trivial tasks.

Employees at Meta created a similar situation. The company’s internal leaderboard, “Claudeonomics,” tracked and ranked individual AI token usage. Within a 30-day window, total AI token usage across the board had exceeded 60 trillion tokens.

Ultimately, in an environment where AI use appears to be the most important metric worth tracking, employees are taking to new extremes to gain company-wide recognition, and potentially even solidify their positions in an increasingly uncertain job market.

The Dangers of Tokenmaxxing for Businesses

Ultimately, using AI token usage as a measure of worker productivity is inherently flawed. For example, an employee can easily burn through AI tokens by running agents continuously through their mailbox to sort messages. Meanwhile, another employee could create an agent that can sort, respond to, and prioritize emails, but with fewer tokens.

Instead of looking at how much AI is being used, businesses should focus on what is being created. This, in turn, will send a positive message to employees about the purpose of AI within your business, and its role as a productivity booster.

Similarly, research by Tech.co shows that companies spend 26% of an hour of AI use reworking output. If employees are purposefully maximizing their use of AI tools, this likely leads to more time spent reviewing, slowing down productivity. In turn, businesses might find they are spending more on AI than they are seeing results.

OpenAI is launching the OpenAI Deployment Company in a push to help enterprises embed AI into their operations. The new company is designed to help organizations build and deploy their own systems to support “their most important work,” according to a statement.

Reportedly, the OpenAI Deployment Company will enable OpenAI to to embed specialized engineers into organizations that are seeking to solve complex issues. Known as Forward Deployed Engineers (FDEs), they will work with business leaders and frontline teams to identify where AI can be deployed.

ChatGPT, OpenAI’s flagship product, is still the de facto AI model of choice for many businesses, but Claude is one of the fastest growing and a favorite of developers.

OpenAI In Push to Help Enterprises Embed AI

OpenAI has announced the launch of a new company, OpenAI Deployment Company, as it gears up to help enterprises embed AI into their systems.

Reportedly, the new initiative will allow OpenAI to deploy more specialized engineers, known as Forward Deployed Engineers (FDEs).

Specializing in “frontier AI deployment,” these engineers enable businesses to solve “complex problems in demanding environments,” according to a statement. They will work in close collaboration with both frontline teams and business leaders to identify areas where AI could be deployed to solve problems.

The new company has launched with $4 billion of investment at a $10 billion pre-money valuation, with backers including Goldman Sach, Bain & Co., Capgemini, and McKinsey & Co.

OpenAI Acquires AI Consulting Firm

OpenAI has also acquired Tomoro, an applied AI consulting and engineering firm. As part of the move, 150 FDEs will be made available to OpenAI Deployment Company, which in turn will help more organizations to embed AI into their workflows.

Ultimately, OpenAI has stated that its decision to launch a new deployment vehicle is in the service of “building for where frontier AI is headed.” Not only will it help new businesses to embed AI into their systems, but it will also operate as an extension of OpenAI, continuing to research, develop, and launch the latest AI technology.

“AI is becoming capable of doing increasingly meaningful work inside organizations. The challenge now is helping companies integrate these systems into the infrastructure and workflows that power their businesses. DeployCo is designed to help organizations bridge that gap and turn AI capability into real operational impact.” – Denise Dresser, OpenAI Chief Revenue Officer

ChatGPT Still Winning the Chatbot Race

Since launching in November 2022, OpenAI’s ChatGPT has enjoyed first-mover advantage and, subsequently, a dominant market share among users and businesses alike. And with adoption increasing at a scarcely believable pace, this trend is unlikely to abate.

More broadly, the AI race continues to unfold at breakneck speed. The recent explosion of different vibe coding platforms has opened up another avenue for users to explore, with tools such as Base44 and Lovable allowing non-technical users to code their own websites and apps through the use of AI.

All of this deployment comes at a cost — more and more, businesses around the world are laying off staff in favor of automation. While many people will try to deny the link between the two phenomena, it’s becoming increasingly difficult to ignore.

Google has just reported the first (known) example of hackers using AI to uncover and exploit a zero-day security flaw.

It’s a sign that the AI threat many experts predicted is officially here for the cybersecurity industry: AI tools can help bad actors discover new pathways towards their goals.

Here are the details to know.

How AI Developed a Zero-Day Exploit

The news comes from a freshly released report by the Google Threat Intelligence Group (GTIG) that follows up on the group’s February 2026 report covering AI-related threat activity.

Now, they say, they have for the very first time “identified a threat actor using a zero-day exploit that we believe was developed with AI.” Specifically, it was within a Python script that let the user bypass two-factor authentication on “a popular open-source, web-based system administration tool.”

The report says that it likely stopped this particular threat:

“The criminal threat actor planned to use it in a mass exploitation event but our proactive counter discovery may have prevented its use. Threat actors associated with the People’s Republic of China (PRC) and the Democratic People’s Republic of Korea (DPRK) have also demonstrated significant interest in capitalizing on AI for vulnerability discovery.” – GTIG

For hackers, the report concludes, AI tools represent “expert-level force multipliers for vulnerability research and exploit development.”

How Google Identified the Use of AI

You might be wondering what makes Google so confident that the hackers relied on AI specifically to uncover the exploit itself – after all, the code didn’t appear to have included the word “delve.”

However, it did include other tell-tale signs, Google says.

“[T]he script contains an abundance of educational docstrings, including a hallucinated CVSS score, and uses a structured, textbook Pythonic format highly characteristic of LLMs training data (e.g., detailed help menus and the clean _C ANSI color class).” -GTIG

The threat was stopped, as GTIG worked with the impacted vendor to disclose the problem.

And don’t worry, Google fans: Their team says that they “we do not believe Gemini was used” for this particular process.

The Age of AI-Powered Cybercrime

AI agents and other LLM-powered tools have been making waves across all areas of business security for years now.

That’s not all good news. Recently, Anthropic’s Cursor wiped a company’s entire database within just a few seconds — before then issuing an apology.

That’s not stopping the rollout of even more AI. At least, not according to a new Gartner prediction that claims the average global Fortune 500 company will have over 150,000 AI agents by 2028.

For cybersecurity professionals, increasing AI use means that criminals and security experts alike have a new set of tools to help each group attempt to one-up the other.

We can only hope the bad actors are at least as likely to accidentally delete their entire database as all the law-abiding professionals.

Anthropic has revealed that its flagship Claude chatbot will soon be able to “dream” in a process mirroring what the human brain does when we sleep. Specifically, it will go over recent events and identify things that are worth committing to “memory,” which can then inform future tasks and interactions.

Unlike compaction — the process that all AI agents undergo in which conversations are regularly audited for irrelevant information — dreaming periodically recurs across multiple agents. This will identify complex patterns that a single agent couldn’t identify alone.

Dreaming is currently limited to Claude Managed Agents, which “provides the harness and infrastructure for running Claude as an autonomous agent,” according to the company. Reportedly, the process will be invaluable for long-running, complex projects that involve multiple AI agents.

Claude Managed Agents Can Soon Dream

Anthropic has revealed that Claude will soon be able to dream. In a process closely mirroring what the human brain does when we sleep, the chatbot will soon revisit recent interactions and identify specific things that are worth storing in its “memory.”

The feature is currently reserved for Claude Managed Agents, which are pre-built agents that operate in managed infrastructure. They’re engineered for long-running and complex tasks that require a high degree of asynchronous work.

Anthropic revealed the news at its latest Code with Claude developers conference. It is still in research preview, but interested developers can request access.

Dreaming Will be Invaluable for Complex Processes

Regular chatbots already undergo a process known as compaction, in which lengthy conversations with users are regularly analyzed, and the models remove information that is irrelevant to the context window. However, Anthropic claims that dreaming differs in a few key ways.

To begin with, while compaction is limited to a single conversation with a single agent, dreaming allows conversations across multiple agents to be analyzed and stored. In theory, this will enable the agents to identify complex patterns. It could be a game changer for developers working on multi-step projects that require a high level of asynchronous work.

“Dreaming surfaces patterns that a single agent can’t see on its own, including recurring mistakes, workflows that agents converge on, and preferences shared across a team. It also restructures memory so it stays high-signal as it evolves.” – Anthropic spokesperson

Claude the Go-To Platform for Developers

From a developer standpoint, Claude continues to go from strength to strength. The platform is already considered the de facto solution for coders everywhere, with the company announcing recently it is rolling out AI agents for financial services.

AI is already deeply embedded in software development, with 90% of developers now using the technology at work, and this trend is showing no signs of abating, with AI shown to accelerate coding. The rollout of various vibe coding platforms has taken this even further.

This comes in spite of various warnings from leading figures in the tech space that mass unemployment due to AI is likely.

Anthropic is rolling out even more AI business tools, with the company announcing new agents that are specifically aimed at improving financial services.

These ready-to-run templates will be able to perform a wide range of tasks, from preparing meetings and building pitches to reviewing valuations and auditing statements.

With the release of these ten AI agents, Anthropic continues to strengthen its enterprise offerings, which have already become the go-to option for businesses in 2026.

Anthropic Launches New AI Agents for Financial Services

Announced in a company blog post, Anthropic unveiled ten new, ready-to-launch AI agent templates that are specifically designed to handle one of the most important aspects of any business.

“Finance is a great blueprint for the rest of knowledge work.” – Nicholas Lin, Anthropic’s head of product for financial services to Bloomberg

There are ten new financial service templates, which can perform and automate a wide range of tasks, including meeting preparation, statement auditing, earnings review, and market research.

How Do Anthropic’s Financial Service Agents Work?

There are two distinct ways in which you can utilize these new AI agents from Anthropic to make financial tasks a lot easier.

For starters, each of the ten agent templates are available as a plug-in for Claude Cowork and Claude Code, with these tools running alongside the analyst to provide more specific assistance depending on the agent.

Additionally, you can use any of these agent templates as a cookbook for Claude Managed Agents, which runs autonomously alongside the Claude platform to streamline your financial operations.

Claude for Business Is Booming

We’ll just come right out and say it: Claude is becoming an absolute powerhouse for businesses and the enterprise industry is taking notice.

The launch of Claude Cowork and Claude Code in January has spurred some serious interest in the AI model, with Anthropic now accounting for 73.3% of new AI spending, compared with only 26.7% from OpenAI.

With these new financial service agent templates, Claude could be positioning itself as the best option for businesses for the foreseeable future.

Check out our Claude for Business guide to learn more