Thales Group Report Reveals Surge in AI Agent Adoption and Rising Cybersecurity Budgets Worldwide
Thales Group has released a comprehensive report examining how artificial intelligence is reshaping information security practices around the globe. The study, based on responses from more than 3,000 participants in 20 countries, highlights both the opportunities and the mounting risks created by widespread AI adoption.
According to the findings, 34 percent of organizations are already using AI agents in their operations, while 73 percent intend to implement them within the coming year. The introduction of these autonomous agents has accelerated data processing speeds and volumes, making it significantly harder for security teams to maintain adequate protection levels.
As a direct consequence, the number of companies earmarking dedicated budgets for AI-related security has grown by 50 percent year-over-year. Last year only 20 percent of respondents allocated specific funds for AI security; this year the figure has risen to 30 percent. Nevertheless, 53 percent still cover AI protection costs from their general information-security budgets.
Speed of change within AI ecosystems was cited by 70 percent of respondents as a major security challenge. Additionally, 61 percent confirmed that their AI applications had become direct targets of attacks aimed at exfiltrating sensitive information. New forms of reputational harm are also emerging: 48 percent of organizations reported financial losses caused by AI-generated disinformation, and 59 percent encountered deepfake-based attacks.
Among realized AI-driven attacks, those focused on disclosing confidential data showed the strongest growth. Cloud resources continue to dominate attacker priorities, with cloud storage named by 35 percent, SaaS applications by 34 percent, and cloud management infrastructure by 32 percent of respondents. User credentials have climbed from eighth to fifth place among top targets, cited by 21 percent, while identity data moved from tenth to sixth place.
Regular security audits demonstrate clear benefits. More than half of organizations that conduct routine audits reported no data leaks from cloud services (54 percent) or corporate systems (57 percent). In contrast, only 6 percent of companies that skipped audits could claim the same. Human error remains the leading cause of incidents, accounting for 28 percent of breaches.
Russian companies share concerns about public AI services
The InfoWatch expert center also presented results from a survey conducted by the Russian BISA association among domestic information-security and IT specialists. Respondents identified the transfer of confidential data to public AI services as the primary threat vector (89 percent), followed by risks associated with embedded AI agents and unintended disclosure to third parties (both 59 percent).
Almost half (46 percent) of Russian respondents said they had either experienced or were aware of data leaks involving generative AI tools. While 59 percent acknowledged the value of AI in daily workflows, they stressed the need for better regulation, methods to detect AI-generated text, and mechanisms to recall data from public models. Most organizations currently allow limited use of AI tools (46 percent) and rely on employee training (48 percent) and enhanced monitoring (30 percent) to mitigate risks.
Related articles
Claude Encrypted Thinking Blocks Use Protobuf with Exposed Metadata and AES-GCM Ciphertext
A detailed reverse-engineering of Claude signatures shows that the encrypted reasoning blocks are not opaque containers but structured protobuf messages. The outer envelope contains a 312-byte inner message that holds a 135-byte header, fixed-length nonce and MAC fields, and the actual ciphertext. The header itself reveals the model name such as claude-opus-5, the block type as thinking, and the organizationUuid from the user's Anthropic account. Only the reasoning text is encrypted with AES-GCM, adding exactly 16 bytes for the authentication tag. The analysis covers four protocol versions and notes that organization binding was added in version 15, potentially allowing servers to reject cross-model or cross-organization reuse. The findings provide concrete implications for both the Opus-to-Haiku extraction attack and the leakage of account identifiers in public logs.
Researchers Extract Proprietary Reasoning Traces from Anthropic, OpenAI and Google LLMs, Revealing Hidden Secrets
A team of eight researchers from institutions including ELLIS Institute Tübingen, the Max Planck Institute for Intelligent Systems, Tübingen AI Center, MATS and Snyk published a preprint detailing a practical attack that recovers full reasoning traces from closed LLM APIs. The method requires only two API calls and works by feeding encrypted reasoning blocks from strong models such as Claude Opus 4.8 into weaker models from the same provider, such as Haiku 4.5, which then reproduce the hidden chain-of-thought verbatim. Analysis of 6,708 publicly shared agent logs from GitHub and Hugging Face yielded 315,320 recovered traces containing 704 unique secrets, including 62 API keys, 33 passwords and 24 access tokens that never appeared in visible session output. The attack also enables extraction of internal safety policies, system prompts and detailed harmful planning that providers normally filter from final answers. In addition, the same mechanism can be used in reverse to inject malicious instructions into shared logs that later get replayed by unsuspecting users. The authors recommend treating encrypted reasoning blocks as sensitive secrets and propose cryptographic binding of traces to sessions, users and models.
GhostSplice Technique Lets Malicious MCP Servers Trick AI Coding Agents into Exfiltrating Secrets
GhostSplice is a new technique that allows a malicious MCP server to induce an AI coding agent to leak SSH keys, environment secrets, and source code. The attack splits malicious instructions across tool metadata and responses so the agent reconstructs and executes the full exfiltration plan without detecting an overtly malicious command. Tests showed the method raised compliance rates from an average of 42 percent to 82 percent across eleven models, with some systems moving from zero to 100 percent success. The technique requires the developer to connect the attacker-controlled MCP server and for the agent to already possess read access to the targeted files. Defenses focus on strict allow-listing of MCP servers, least-privilege tool permissions, separation of tool output from instructions, and human approval for sensitive operations. The disclosure aligns with prior warnings about poisoned MCP tool descriptions and agentjacking attacks.
Anthropic Claude Code Auto Mode Launches August 14 with Local Classifier and Permission Rules
Starting August 14, Claude Code will run in auto mode on new sessions for Pro, Max, and Team plans, replacing the allow/deny dialog with a local classifier that evaluates every tool call. The classifier rules are stored locally and contain 103 categories across allow, soft_deny, hard_deny, and environment sections, with the single hard_deny rule focused on data exfiltration spanning over 5,000 characters. Enterprise, API, Bedrock, Vertex, and Foundry deployments remain on opt-in for another month. Auto mode pauses after three consecutive blocks or twenty blocks in a session, and broad allow rules such as python:* are disabled while narrow permissions continue to function. Administrators should populate the twenty environment fields, currently only one-third configured on clean machines, before the rollout date.