BSI and TÜV: AI attacks hit one in six firms
3 min read
17 percent of surveyed companies with at least 20 employees recorded confirmed or suspected AI-assisted cyberattacks or fraud attempts in the past twelve months. Only 11 percent of businesses using or planning to use AI have dedicated emergency procedures for AI-related security incidents. Companies without their own AI are also at risk, warns BSI (Federal Office for Information Security).
Key takeaways
- 17 percent of surveyed companies report AI-assisted attacks or fraud attempts. 4 percent report only confirmed incidents, 11 percent only suspected incidents and 2 percent both.
- Only 11 percent of businesses using or planning to use AI have dedicated emergency procedures. 43 percent rely on general IT security procedures, while 25 percent have no rules in place.
- For urgent payment and data release requests, a callback through a known channel is recommended.
Related:Neuralink restores an ALS patient’s voice / Bitkom: 90 percent of firms want GDPR reform
Confirmed and suspected incidents
The findings of the joint 2026 AI security study by BSI and TÜV-Verband (TÜV Association) have been available since 7 October. They are based on a survey of 505 companies with at least 20 employees, conducted by research institute forsa between 22 June and 31 July 2026. 17 percent of these companies recorded AI-assisted cyberattacks or fraud attempts in the preceding twelve months.
The figure includes findings with different levels of certainty. 4 percent of companies report only confirmed incidents, 11 percent only suspected incidents and 2 percent both. Dr Dirk Stenkamp, president of TÜV-Verband, warned that the AI-assisted attacks companies have detected so far could be just the tip of the iceberg.
What is a deepfake? A deepfake is a fake audio or video recording generated using AI. Attackers use these fakes, for example, to imitate an executive’s voice. 11 percent of companies that recorded AI-assisted attacks cited deepfakes as an attack method.
Phishing leads the list
Among affected companies, highly realistic phishing emails were the most common attack method, at 90 percent. Phishing involves fake messages designed, for example, to trick recipients into disclosing login credentials. Automated attack scripts followed at 40 percent. These are programs that search for vulnerabilities autonomously.
Gaps in emergency procedures
Among companies using or planning to use AI, only 11 percent have dedicated procedures for AI-related security incidents. These procedures define who makes decisions, communicates and takes action during an incident. 43 percent rely on general IT security procedures without AI-specific rules. 20 percent are developing such procedures, while 25 percent have no rules in place.
Only 31 percent of companies using AI deploy it to improve their own IT security. Stenkamp recommended adding AI-related scenarios to existing security and emergency procedures.
Risks extend to companies without AI
49 percent of surveyed companies use AI, while another 9 percent plan to adopt it within the next twelve months. 42 percent do neither. BSI vice president Thomas Caspers noted that companies that do not use AI themselves are also at risk from AI-assisted cyberattacks. He urged all businesses to promptly consider AI as a tool for their own cybersecurity.
Immediate steps for small businesses
A handout for small business owners has been published alongside the study by the Alliance for Cyber Security (ACS), based at BSI. Entitled “Protection against AI-enhanced cyberattacks”, it provides practical recommendations and immediate steps. Unusual, urgent requests to authorise payments or release data should be verified through a callback initiated by the recipient using another known, secure communication channel. Binding rules for AI use within the business are also recommended.
Frequently asked questions
Who took part in the AI security study?
The survey covered staff responsible for AI use, IT security or IT management, as well as senior management, at 505 companies with at least 20 employees. It used computer-assisted telephone interviewing (CATI). forsa conducted the survey on behalf of BSI and TÜV-Verband.
Why do 42 percent of companies not use AI?
These businesses most frequently cite a lack of benefits (58 percent) and data protection or security concerns (53 percent). High integration effort follows at 44 percent, with legal uncertainties at 43 percent.
When will certification for AI systems be available?
The TÜV organisations are developing a framework for certifying AI systems together with TÜV AI.Lab. Pilot partners are testing it, and it is due to be available at the end of 2026. 69 percent of companies using or planning to use AI expect independent assessments to improve security and reliability. For 51 percent, certification is an important selection criterion.
Read more on MyBusinessFuture
MyBusinessFutureE-invoicing: Three decisions for the MittelstandMyBusinessFutureIndustry in August: Major orders fall away, core business holds upMyBusinessFutureOne year of mandatory electronic patient records: Connected but rarely usedMore from the MBF Media Network
cloudmagazinMistral brings Europe closer to the AI leadersDigital ChiefsAI spending rises; only one in five firms has scaled itSecurityTodayCyber insurance: lack of evidence puts the policy at riskImage credit: AI-generated (October 2026)
Image source: AI-generated (October 2026)
Translated from the German original using artificial intelligence. The German version is authoritative.
Also available in
